Hello all.
If we don't install libreoffice by default we'd save around ~500MB (of libreoffice itself and it's various java dependencies) from the live media. This means less strain on mirrors, and that people will get their media faster (even if you have a really fast connection - smaller downloads finish sooner).
Also, since we have an application installer these days, people who need an office suite can easily install it.
I suggest we remove it from the default install.
Almost everything can be installed after the fact. Media players, web browsers, document viewers, games. None are "required" and can be installed afterwards.
That does not mean that an office suite isnt useful.
Having an office suite installed gives the impression that Fedora is serious and redy for work, and also stops the default installation from being seen as a "toy".
When OOO.o wasnt installed by default, it used to be one of the first steps carried out by me (and I suspect many others). Having it on the live media is IMO a good thing.
On 12 September 2014 11:06, Elad Alfassa elad@fedoraproject.org wrote:
Hello all.
If we don't install libreoffice by default we'd save around ~500MB (of libreoffice itself and it's various java dependencies) from the live media. This means less strain on mirrors, and that people will get their media faster (even if you have a really fast connection - smaller downloads finish sooner).
Also, since we have an application installer these days, people who need an office suite can easily install it.
I suggest we remove it from the default install.
--
-Elad Alfassa.
desktop mailing list desktop@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/desktop
Hello,
I agree with Elad Alfassa, Libre Office makes the images huge (what about the net-install image?) and it also affects updates. When someone gets the whole Libre Office suite by default, he may not never use it, but he may not ever remove it, and that leads to bigger updates.
LibO can be easy discover-able through Software or even Software search provider -at least in GNOME desktop-, and over and above LibO is a very famous project that everyone knows it anyway.
About how useful/popular is for the users ..who knows? I personally prefer Office on Cloud services.
- alex
On Fri, Sep 12, 2014 at 1:12 PM, Naheem Zaffar naheemzaffar@gmail.com wrote:
Almost everything can be installed after the fact. Media players, web browsers, document viewers, games. None are "required" and can be installed afterwards.
That does not mean that an office suite isnt useful.
Having an office suite installed gives the impression that Fedora is serious and redy for work, and also stops the default installation from being seen as a "toy".
When OOO.o wasnt installed by default, it used to be one of the first steps carried out by me (and I suspect many others). Having it on the live media is IMO a good thing.
On 12 September 2014 11:06, Elad Alfassa elad@fedoraproject.org wrote:
Hello all.
If we don't install libreoffice by default we'd save around ~500MB (of libreoffice itself and it's various java dependencies) from the live media. This means less strain on mirrors, and that people will get their media faster (even if you have a really fast connection - smaller downloads finish sooner).
Also, since we have an application installer these days, people who need an office suite can easily install it.
I suggest we remove it from the default install.
--
-Elad Alfassa.
desktop mailing list desktop@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/desktop
-- desktop mailing list desktop@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/desktop
I suspect that while the technical merits of removing LO from a default install have their place (ie. large image size, quicker downloads, less update server load, it can be easily discovered in and installed from the repos, there are popular alternatives like cloud suites and alternatives, etc. etc. etc.) it's a major set of packages, *for me*, and I suspect for a significant proportion of users, even if a minority (in which case I concede that that would be a further reason to consider its removal.)
Oh, here's one more technical reason to remove it, or at least modify how it's installed: Although I know that there is a common codebase to the various parts, what about installing LO piece-meal? I use Write and Calc all the time. I have used Impress in the past but in the past three or four years I've used it perhaps three times. I've never used Base, Draw, Math or Charts since starting to use OpenOffice.org in 2005.
Beyond the question of "serious and ready for work" impression its inclusion provides, I'd say that if it is removed, then another (suite of) "killer app(s)" needs to be showcased. Unfortunately, a rock-solid base full of "a few nice apps" isn't good enough. Note that *for*me* that "killer app" is LO. For someone else, LO definitely *isn't*. For me the Security Spin is an example of what I mean: (1) lists all the major packages. I don't see LO in the list, but I consider the spin to be "useful out of the box" despite the ommission, because LO isn't what the Spin is about.
So I'm suggesting that I probably won't publicly care either way (and mumble while installing LO manually, were I to use the WS live iso to install) but that from a marketing and public image perspective, the ISO should be "complete" -- whatever that means -- out of the box.
note: I don't like installing from live images, for me it's all or nothing. I've been using the "full fat" iso or increasingly the net-install iso for years. Further, I'm in the camp of "easily able to navigate and install from the repos on my own in order to start customizing my setup".
(1) https://fedorahosted.org/security-spin/wiki/availableApps
Hello,
I agree with Elad Alfassa, Libre Office makes the images huge (what about the net-install image?) and it also affects updates. When someone gets the whole Libre Office suite by default, he may not never use it, but he may not ever remove it, and that leads to bigger updates.
LibO can be easy discover-able through Software or even Software search provider -at least in GNOME desktop-, and over and above LibO is a very famous project that everyone knows it anyway.
About how useful/popular is for the users ..who knows? I personally prefer Office on Cloud services.
- alex
On Fri, Sep 12, 2014 at 1:12 PM, Naheem Zaffar naheemzaffar@gmail.com wrote:
Almost everything can be installed after the fact. Media players, web browsers, document viewers, games. None are "required" and can be installed afterwards.
That does not mean that an office suite isnt useful.
Having an office suite installed gives the impression that Fedora is serious and redy for work, and also stops the default installation from being seen as a "toy".
When OOO.o wasnt installed by default, it used to be one of the first steps carried out by me (and I suspect many others). Having it on the live media is IMO a good thing.
On 12 September 2014 11:06, Elad Alfassa elad@fedoraproject.org wrote:
Hello all.
If we don't install libreoffice by default we'd save around ~500MB (of libreoffice itself and it's various java dependencies) from the live media. This means less strain on mirrors, and that people will get their media faster (even if you have a really fast connection - smaller downloads finish sooner).
Also, since we have an application installer these days, people who need an office suite can easily install it.
I suggest we remove it from the default install.
--
-Elad Alfassa.
desktop mailing list desktop@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/desktop
-- desktop mailing list desktop@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/desktop
-- desktop mailing list desktop@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/desktop
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 09/12/2014 08:27 AM, Donald Buchan wrote:
I suspect that while the technical merits of removing LO from a default install have their place (ie. large image size, quicker downloads, less update server load, it can be easily discovered in and installed from the repos, there are popular alternatives like cloud suites and alternatives, etc. etc. etc.) it's a major set of packages, *for me*, and I suspect for a significant proportion of users, even if a minority (in which case I concede that that would be a further reason to consider its removal.)
- From my perspective, the point of offering a Live Media instead of a pure installer is so that people can run Fedora and do something actually useful on it without having to clobber their existing system at first. (The try-before-you-buy scenario)
We really need to figure out (ideally through user testing), what are the tasks that people would want to do before they install Fedora to their local system. I'd strongly argue that the most basic set of tasks would be:
1) Browse the internet 2) Check my email 3) Read/Write office documents 4) Use instant messaging (Google Hangouts, Facebook messenger)
For Fedora contributors, I might also add 4) Connect to IRC
Oh, here's one more technical reason to remove it, or at least modify how it's installed: Although I know that there is a common codebase to the various parts, what about installing LO piece-meal? I use Write and Calc all the time. I have used Impress in the past but in the past three or four years I've used it perhaps three times. I've never used Base, Draw, Math or Charts since starting to use OpenOffice.org in 2005.
I just performed a little test. Removing Base, Draw, Math and Charts from my installed machine saves a whopping 6.3MB. Even if we try to account for a few different dependencies, the savings are negligible.
The vast majority of disk space is taken up by libreoffice-core (241MB on its own) and its dependencies.
----- Original Message -----
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 09/12/2014 08:27 AM, Donald Buchan wrote:
I suspect that while the technical merits of removing LO from a default install have their place (ie. large image size, quicker downloads, less update server load, it can be easily discovered in and installed from the repos, there are popular alternatives like cloud suites and alternatives, etc. etc. etc.) it's a major set of packages, *for me*, and I suspect for a significant proportion of users, even if a minority (in which case I concede that that would be a further reason to consider its removal.)
- From my perspective, the point of offering a Live Media instead of a
pure installer is so that people can run Fedora and do something actually useful on it without having to clobber their existing system at first. (The try-before-you-buy scenario)
We really need to figure out (ideally through user testing), what are the tasks that people would want to do before they install Fedora to their local system. I'd strongly argue that the most basic set of tasks would be:
- Browse the internet
- Check my email
- Read/Write office documents
- Use instant messaging (Google Hangouts, Facebook messenger)
For Fedora contributors, I might also add 4) Connect to IRC
Oh, here's one more technical reason to remove it, or at least modify how it's installed: Although I know that there is a common codebase to the various parts, what about installing LO piece-meal? I use Write and Calc all the time. I have used Impress in the past but in the past three or four years I've used it perhaps three times. I've never used Base, Draw, Math or Charts since starting to use OpenOffice.org in 2005.
I just performed a little test. Removing Base, Draw, Math and Charts from my installed machine saves a whopping 6.3MB. Even if we try to account for a few different dependencies, the savings are negligible.
The vast majority of disk space is taken up by libreoffice-core (241MB on its own) and its dependencies.
If you add the dependencies that LibreOffice drags (and wouldn't otherwise be on the live CD), you can add another 250 megs.
On Fri, 2014-09-12 at 08:39 -0400, Stephen Gallagher wrote:
On 09/12/2014 08:27 AM, Donald Buchan wrote:
I suspect that while the technical merits of removing LO from a default install have their place (ie. large image size, quicker downloads, less update server load, it can be easily discovered in and installed from the repos, there are popular alternatives like cloud suites and alternatives, etc. etc. etc.) it's a major set of packages, *for me*, and I suspect for a significant proportion of users, even if a minority (in which case I concede that that would be a further reason to consider its removal.)
From my perspective, the point of offering a Live Media instead of a pure installer is so that people can run Fedora and do something actually useful on it without having to clobber their existing system at first. (The try-before-you-buy scenario)
Certainly from the "try before you buy" scenario then I personally would wonder ( http://www.malak.ca/blog/?p=227 ) why it would be missing. Someone else, maybe not, who knows? (Hence why I think that a killer app or suite of apps would still be needed, regardless of what they are.)
We really need to figure out (ideally through user testing), what are the tasks that people would want to do before they install Fedora to their local system.
(snip)
I was starting to think that after I'd sent my message.
And if LO proves to be wanted by a sufficient minority and removing it would be considered "a good thing", then as I said I personally would have no difficulty installing it on my own anyway. (In any case, I'm a more general-use case user and don't fall into the target developper audience, and I have sufficient technical savvy to tailor my system to what I want it to do.)
Oh, here's one more technical reason to remove it, or at least modify how it's installed: Although I know that there is a common codebase to the various parts, what about installing LO piece-meal? I use Write and Calc all the time. I have used Impress in the past but in the past three or four years I've used it perhaps three times. I've never used Base, Draw, Math or Charts since starting to use OpenOffice.org in 2005.
I just performed a little test. Removing Base, Draw, Math and Charts from my installed machine saves a whopping 6.3MB. Even if we try to account for a few different dependencies, the savings are negligible.
The vast majority of disk space is taken up by libreoffice-core (241MB on its own) and its dependencies.
Thanks!
On Fri, Sep 12, 2014 at 08:39:54AM -0400, Stephen Gallagher wrote:
On 09/12/2014 08:27 AM, Donald Buchan wrote:
I suspect that while the technical merits of removing LO from a default install have their place (ie. large image size, quicker downloads, less update server load, it can be easily discovered in and installed from the repos, there are popular alternatives like cloud suites and alternatives, etc. etc. etc.) it's a major set of packages, *for me*, and I suspect for a significant proportion of users, even if a minority (in which case I concede that that would be a further reason to consider its removal.)
- From my perspective, the point of offering a Live Media instead of a
pure installer is so that people can run Fedora and do something actually useful on it without having to clobber their existing system at first. (The try-before-you-buy scenario)
We really need to figure out (ideally through user testing), what are the tasks that people would want to do before they install Fedora to their local system. I'd strongly argue that the most basic set of tasks would be:
- Browse the internet
- Check my email
- Read/Write office documents
- Use instant messaging (Google Hangouts, Facebook messenger)
For Fedora contributors, I might also add 4) Connect to IRC
Oh, here's one more technical reason to remove it, or at least modify how it's installed: Although I know that there is a common codebase to the various parts, what about installing LO piece-meal? I use Write and Calc all the time. I have used Impress in the past but in the past three or four years I've used it perhaps three times. I've never used Base, Draw, Math or Charts since starting to use OpenOffice.org in 2005.
I just performed a little test. Removing Base, Draw, Math and Charts from my installed machine saves a whopping 6.3MB. Even if we try to account for a few different dependencies, the savings are negligible.
The vast majority of disk space is taken up by libreoffice-core (241MB on its own) and its dependencies.
The utility argument seems pretty strong to me. Also, I believe deltarpm is still enabled by default in installations, correct? That would tend to cut down on size of updates to download.
On Fri, Sep 12, 2014 at 4:26 PM, Paul W. Frields stickster@gmail.com wrote:
The utility argument seems pretty strong to me. Also, I believe deltarpm is still enabled by default in installations, correct? That would tend to cut down on size of updates to download.
I'm not sure if PackageKit's hif backend can do deltarpms.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On 09/12/2014 09:26 AM, Paul W. Frields wrote:
On Fri, Sep 12, 2014 at 08:39:54AM -0400, Stephen Gallagher wrote:
On 09/12/2014 08:27 AM, Donald Buchan wrote:
I suspect that while the technical merits of removing LO from a default install have their place (ie. large image size, quicker downloads, less update server load, it can be easily discovered in and installed from the repos, there are popular alternatives like cloud suites and alternatives, etc. etc. etc.) it's a major set of packages, *for me*, and I suspect for a significant proportion of users, even if a minority (in which case I concede that that would be a further reason to consider its removal.)
- From my perspective, the point of offering a Live Media instead
of a pure installer is so that people can run Fedora and do something actually useful on it without having to clobber their existing system at first. (The try-before-you-buy scenario)
We really need to figure out (ideally through user testing), what are the tasks that people would want to do before they install Fedora to their local system. I'd strongly argue that the most basic set of tasks would be:
- Browse the internet 2) Check my email 3) Read/Write office
documents 4) Use instant messaging (Google Hangouts, Facebook messenger)
For Fedora contributors, I might also add 4) Connect to IRC
Oh, here's one more technical reason to remove it, or at least modify how it's installed: Although I know that there is a common codebase to the various parts, what about installing LO piece-meal? I use Write and Calc all the time. I have used Impress in the past but in the past three or four years I've used it perhaps three times. I've never used Base, Draw, Math or Charts since starting to use OpenOffice.org in 2005.
I just performed a little test. Removing Base, Draw, Math and Charts from my installed machine saves a whopping 6.3MB. Even if we try to account for a few different dependencies, the savings are negligible.
The vast majority of disk space is taken up by libreoffice-core (241MB on its own) and its dependencies.
The utility argument seems pretty strong to me. Also, I believe deltarpm is still enabled by default in installations, correct? That would tend to cut down on size of updates to download.
Deltarpms help only for updates. No matter what, you need to download the full package at least the first time.
On Fri, Sep 12, 2014 at 09:50:21AM -0400, Stephen Gallagher wrote:
On 09/12/2014 09:26 AM, Paul W. Frields wrote:
On Fri, Sep 12, 2014 at 08:39:54AM -0400, Stephen Gallagher wrote:
On 09/12/2014 08:27 AM, Donald Buchan wrote:
I suspect that while the technical merits of removing LO from a default install have their place (ie. large image size, quicker downloads, less update server load, it can be easily discovered in and installed from the repos, there are popular alternatives like cloud suites and alternatives, etc. etc. etc.) it's a major set of packages, *for me*, and I suspect for a significant proportion of users, even if a minority (in which case I concede that that would be a further reason to consider its removal.)
- From my perspective, the point of offering a Live Media instead
of a pure installer is so that people can run Fedora and do something actually useful on it without having to clobber their existing system at first. (The try-before-you-buy scenario)
We really need to figure out (ideally through user testing), what are the tasks that people would want to do before they install Fedora to their local system. I'd strongly argue that the most basic set of tasks would be:
- Browse the internet 2) Check my email 3) Read/Write office
documents 4) Use instant messaging (Google Hangouts, Facebook messenger)
For Fedora contributors, I might also add 4) Connect to IRC
Oh, here's one more technical reason to remove it, or at least modify how it's installed: Although I know that there is a common codebase to the various parts, what about installing LO piece-meal? I use Write and Calc all the time. I have used Impress in the past but in the past three or four years I've used it perhaps three times. I've never used Base, Draw, Math or Charts since starting to use OpenOffice.org in 2005.
I just performed a little test. Removing Base, Draw, Math and Charts from my installed machine saves a whopping 6.3MB. Even if we try to account for a few different dependencies, the savings are negligible.
The vast majority of disk space is taken up by libreoffice-core (241MB on its own) and its dependencies.
The utility argument seems pretty strong to me. Also, I believe deltarpm is still enabled by default in installations, correct? That would tend to cut down on size of updates to download.
Deltarpms help only for updates. No matter what, you need to download the full package at least the first time.
Of course, so the image is necessarily larger as a result. But I question whether the difference saved by eliminating LO would really impact people's ability to download. I'd expect someone who can't download a 1.2GB image wouldn't have their problem solved by only downloading a ~750-950MB image.
On Fri, Sep 12, 2014 at 6:08 PM, Paul W. Frields stickster@gmail.com wrote:
Of course, so the image is necessarily larger as a result. But I question whether the difference saved by eliminating LO would really impact people's ability to download. I'd expect someone who can't download a 1.2GB image wouldn't have their problem solved by only downloading a ~750-950MB image.
Downloading a smaller image is cheaper (if you pay per byte) and it's faster. I, for example, have to wait an hour and a half to download a workstation compose these days, some of the composes being 1.4GB.
On Fri, Sep 12, 2014 at 5:15 PM, Elad Alfassa elad@fedoraproject.org wrote:
On Fri, Sep 12, 2014 at 6:08 PM, Paul W. Frields stickster@gmail.com wrote:
Of course, so the image is necessarily larger as a result. But I question whether the difference saved by eliminating LO would really impact people's ability to download. I'd expect someone who can't download a 1.2GB image wouldn't have their problem solved by only downloading a ~750-950MB image.
Downloading a smaller image is cheaper (if you pay per byte) and it's faster. I, for example, have to wait an hour and a half to download a workstation compose these days, some of the composes being 1.4GB.
Yeah but you probably download it more often than the average user that just does it to install the os and not test composes etc.
----- Original Message -----
On Fri, Sep 12, 2014 at 6:08 PM, Paul W. Frields < stickster@gmail.com > wrote:
Of course, so the image is necessarily larger as a result. But I question whether the difference saved by eliminating LO would really impact people's ability to download. I'd expect someone who can't download a 1.2GB image wouldn't have their problem solved by only downloading a ~750-950MB image.
Downloading a smaller image is cheaper (if you pay per byte) and it's faster. I, for example, have to wait an hour and a half to download a workstation compose these days, some of the composes being 1.4GB.
That's the reason, why we did installation DVDs and we shipped it to countries with limited internet access - it contained almost everything we had and especially office suite. So if we are going to continue to produce media, this actually helps to have OO in the compose.
For saving download - I'd say at this time, almost everyone is going to install LO after installation. Other option would be Calligra, it could be smaller but it's still not on par with LO, even it's becoming better.
Personally, I hope for libre Google Docs like service. WebODF seems to be getting close.
Jaroslav
-- -Elad Alfassa.
-- desktop mailing list desktop@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/desktop
Stephen Gallagher píše v Pá 12. 09. 2014 v 08:39 -0400:
On 09/12/2014 08:27 AM, Donald Buchan wrote:
I suspect that while the technical merits of removing LO from a default install have their place (ie. large image size, quicker downloads, less update server load, it can be easily discovered in and installed from the repos, there are popular alternatives like cloud suites and alternatives, etc. etc. etc.) it's a major set of packages, *for me*, and I suspect for a significant proportion of users, even if a minority (in which case I concede that that would be a further reason to consider its removal.)
From my perspective, the point of offering a Live Media instead of a pure installer is so that people can run Fedora and do something actually useful on it without having to clobber their existing system at first. (The try-before-you-buy scenario)
We really need to figure out (ideally through user testing), what are the tasks that people would want to do before they install Fedora to their local system. I'd strongly argue that the most basic set of tasks would be:
- Browse the internet
- Check my email
- Read/Write office documents
- Use instant messaging (Google Hangouts, Facebook messenger)
For Fedora contributors, I might also add 4) Connect to IRC
LibreOffice was not included in the desktop spin until Fedora 19 (?) and I remember the cries from users that one of the most used programs was not included in the default installation. Although I do see the technical merits of removing LibreOffice I still think that most users consider it as a part of basic software selection that should be pre-installed. Any ideas how other distributions handle this? Whether Ubuntu, openSUSE, Mageia... have LibreOffice pre-installed?
Jiri
On Fri, 2014-09-12 at 08:27 -0400, Donald Buchan wrote:
Beyond the question of "serious and ready for work" impression its inclusion provides
I'm sympathetic to both sides of this discussion, since it would be easy to feature LO in the software center and would save a considerable amount of space, but I think the advantages of coming with a full-fledged office suite pre-included (at least Writer and Impress, and probably also Calc) probably outweigh the downsides. One of the biggest reasons users will hesitate to switch from Windows to Fedora is "can it run Office?" Well, it comes with an office suite. It's a workstation.
F21 will be a great release either way.
Michael
On Fri, Sep 12, 2014 at 08:54:15AM -0500, Michael Catanzaro wrote:
On Fri, 2014-09-12 at 08:27 -0400, Donald Buchan wrote:
Beyond the question of "serious and ready for work" impression its inclusion provides
I'm sympathetic to both sides of this discussion, since it would be easy to feature LO in the software center and would save a considerable amount of space, but I think the advantages of coming with a full-fledged office suite pre-included (at least Writer and Impress, and probably also Calc) probably outweigh the downsides. One of the biggest reasons users will hesitate to switch from Windows to Fedora is "can it run Office?" Well, it comes with an office suite. It's a workstation.
I agree with this FWIW.
F21 will be a great release either way.
This too! :-)
I like that it's included. However, LibreOffice apparently can't be removed in Software. It's not listed in the Installed tab, while it's installed parts are listed in the Updates tab. This is after a TC6 installation. Bug?
And if it can be removed, is it going to remove libreoffice-core? That's kinda important seeing as if all the LibreOffice applications are removed, but core stays behind and continues to be updated, makes it pointless to be able to remove LibreOffice from Software.
Chris Murphy
On Fri, 2014-09-12 at 11:25 -0600, Chris Murphy wrote:
I like that it's included. However, LibreOffice apparently can't be removed in Software. It's not listed in the Installed tab, while it's installed parts are listed in the Updates tab. This is after a TC6 installation. Bug?
Bug. Richard?
And if it can be removed, is it going to remove libreoffice-core? That's kinda important seeing as if all the LibreOffice applications are removed, but core stays behind and continues to be updated, makes it pointless to be able to remove LibreOffice from Software.
I think the answer is no. You're hoping for the equivalent of 'yum autoremove'. gpk-application was able to do that with a hidden dconf option, but I don't think gnome-software has it.
On Sep 12, 2014, at 1:25 PM, Michael Catanzaro mcatanzaro@gnome.org wrote:
On Fri, 2014-09-12 at 11:25 -0600, Chris Murphy wrote:
I like that it's included. However, LibreOffice apparently can't be removed in Software. It's not listed in the Installed tab, while it's installed parts are listed in the Updates tab. This is after a TC6 installation. Bug?
Bug. Richard?
And if it can be removed, is it going to remove libreoffice-core? That's kinda important seeing as if all the LibreOffice applications are removed, but core stays behind and continues to be updated, makes it pointless to be able to remove LibreOffice from Software.
I think the answer is no. You're hoping for the equivalent of 'yum autoremove'. gpk-application was able to do that with a hidden dconf option, but I don't think gnome-software has it.
Too bad. Well, I'd hem and haw and probably end up very slightly on the side of not installing it by default just because it's cleaner to not always be updating something that's not being used. If there were some usage/install statistics an objective decision would be possible, rather than just having a personal opinion.
What happens in the installer UI if LibreOffice isn't on the media? Does it still show up as an Add On, and therefore get downloaded and installed in that case? Or is it also delisted as an Add On?
Chris Murphy
On 15 September 2014 02:48, Chris Murphy lists@colorremedies.com wrote:
What happens in the installer UI if LibreOffice isn't on the media? Does it still show up as an Add On, and therefore get downloaded and installed in that case? Or is it also delisted as an Add On?
We don't show any difference in the UI for available-locally and available-remotely applications, but I may have misunderstood your question.
Richard
On Sep 16, 2014, at 2:46 AM, Richard Hughes hughsient@gmail.com wrote:
On 15 September 2014 02:48, Chris Murphy lists@colorremedies.com wrote:
What happens in the installer UI if LibreOffice isn't on the media? Does it still show up as an Add On, and therefore get downloaded and installed in that case? Or is it also delisted as an Add On?
We don't show any difference in the UI for available-locally and available-remotely applications, but I may have misunderstood your question.
You didn't. I spaced out Lives don't have "Software Selection" at all in the installer UI.
Chris Murphy
A few things worth considering:
- Java is quite popular among the Workstation target audience (between web and mobile development), so installing it by default may not be such a bad idea.
- Without Libreoffice there's no support for .doc or even RTF documents in the default install. This is something all operating systems currently support (Windows, OS X and all major Linux distributions). While I'm in favor of offering a lean selection of apps now that Software is a great and usable tool, I believe that an OS should support viewing all common MIME types out of the box.
-- Evandro
On Fri, Sep 12, 2014 at 7:06 AM, Elad Alfassa elad@fedoraproject.org wrote:
Hello all.
If we don't install libreoffice by default we'd save around ~500MB (of libreoffice itself and it's various java dependencies) from the live media. This means less strain on mirrors, and that people will get their media faster (even if you have a really fast connection - smaller downloads finish sooner).
Also, since we have an application installer these days, people who need an office suite can easily install it.
I suggest we remove it from the default install.
--
-Elad Alfassa.
desktop mailing list desktop@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/desktop
On Mon, Sep 15, 2014 at 11:40 PM, Evandro Giovanini efgiovanini@gmail.com wrote:
A few things worth considering:
- Java is quite popular among the Workstation target audience (between web
and mobile development), so installing it by default may not be such a bad idea.
- Without Libreoffice there's no support for .doc or even RTF documents in
the default install. This is something all operating systems currently support (Windows, OS X and all major Linux distributions). While I'm in favor of offering a lean selection of apps now that Software is a great and usable tool, I believe that an OS should support viewing all common MIME types out of the box.
-- Evandr
Having a Java runtime environment doesn't help you with mobile development, you'd have to install Eclipse as well - and that pull in Java anyway. I also don't think a technology being "popular" is a good reason to include it by default, otherwise we'd include ruby on rails, django, docker, and PHP by default.
AFAIK, Microsoft Windows does not come with a .doc viewer by default. In Windows you have to download the viewer separately. I don't know about Mac. Also, when you try to open a file you don't have an app to handle you'd get a dialog asking you if you want to search and install software to open your file.
On Tue, Sep 16, 2014 at 10:54 AM, Elad Alfassa elad@fedoraproject.org wrote:
On Mon, Sep 15, 2014 at 11:40 PM, Evandro Giovanini efgiovanini@gmail.com wrote:
A few things worth considering:
- Java is quite popular among the Workstation target audience (between web
and mobile development), so installing it by default may not be such a bad idea.
- Without Libreoffice there's no support for .doc or even RTF documents in
the default install. This is something all operating systems currently support (Windows, OS X and all major Linux distributions). While I'm in favor of offering a lean selection of apps now that Software is a great and usable tool, I believe that an OS should support viewing all common MIME types out of the box.
-- Evandr
Having a Java runtime environment doesn't help you with mobile development, you'd have to install Eclipse as well - and that pull in Java anyway. I also don't think a technology being "popular" is a good reason to include it by default, otherwise we'd include ruby on rails, django, docker, and PHP by default.
AFAIK, Microsoft Windows does not come with a .doc viewer by default.
It does. Its called "WordPad".
On Sep 16, 2014, at 3:47 AM, drago01 drago01@gmail.com wrote:
On Tue, Sep 16, 2014 at 10:54 AM, Elad Alfassa elad@fedoraproject.org wrote:
On Mon, Sep 15, 2014 at 11:40 PM, Evandro Giovanini efgiovanini@gmail.com wrote:
A few things worth considering:
- Java is quite popular among the Workstation target audience (between web
and mobile development), so installing it by default may not be such a bad idea.
- Without Libreoffice there's no support for .doc or even RTF documents in
the default install. This is something all operating systems currently support (Windows, OS X and all major Linux distributions). While I'm in favor of offering a lean selection of apps now that Software is a great and usable tool, I believe that an OS should support viewing all common MIME types out of the box.
-- Evandr
Having a Java runtime environment doesn't help you with mobile development, you'd have to install Eclipse as well - and that pull in Java anyway. I also don't think a technology being "popular" is a good reason to include it by default, otherwise we'd include ruby on rails, django, docker, and PHP by default.
AFAIK, Microsoft Windows does not come with a .doc viewer by default.
It does. Its called "WordPad".
On OS X it's TextEdit which reads Word doc/docx/xml, .rtf, and .odf. There's intrinsic value having at least a viewer for common formats. If LibreOffice gets us that capability, keep it installed by default.
Chris Murphy
Chris Murphy (lists@colorremedies.com) said:
It does. Its called "WordPad".
On OS X it's TextEdit which reads Word doc/docx/xml, .rtf, and .odf. There's intrinsic value having at least a viewer for common formats. If LibreOffice gets us that capability, keep it installed by default.
gnome-documents would be that interface... which works by using unoconv and LibreOffice.
Bill
Bill Nottingham píše v Út 16. 09. 2014 v 12:19 -0400:
Chris Murphy (lists@colorremedies.com) said:
It does. Its called "WordPad".
On OS X it's TextEdit which reads Word doc/docx/xml, .rtf, and .odf. There's intrinsic value having at least a viewer for common formats. If LibreOffice gets us that capability, keep it installed by default.
gnome-documents would be that interface... which works by using unoconv and LibreOffice.
The current document viewer in GNOME Documents is not even remotely on par with LibreOffice. Until GNOME Documents starts using liblibreoffice, which is planned, I would not consider it as a sufficient option.
Jiri
On 09/12/2014 06:06 AM, Elad Alfassa wrote:
Hello all.
If we don't install libreoffice by default we'd save around ~500MB (of libreoffice itself and it's various java dependencies) from the live media. This means less strain on mirrors, and that people will get their media faster (even if you have a really fast connection - smaller downloads finish sooner).
Also, since we have an application installer these days, people who need an office suite can easily install it.
The ease of finding and installing an office suite when there isn't one installed by default would be something that would make a great user test scenario before yanking libreoffice.
Both the scenarios of opening an exisiting file, and trying to create a document -- some of the things to consider could be: * what people search for -- (do all users use "Word Processor", or just "Word") * if they search for a word processor in Software, does LO show up as a "best bet" or do other applications like abiword or calligra show up? -- a default here may help the user because they are getting the software that may be considered "the best"
cheers, ryanlerch
Hi,
On 09/16/2014 12:18 PM, Ryan Lerch wrote:
The ease of finding and installing an office suite when there isn't one installed by default would be something that would make a great user test scenario before yanking libreoffice.
Both the scenarios of opening an exisiting file, and trying to create a document -- some of the things to consider could be:
- what people search for -- (do all users use "Word Processor", or just
"Word")
- if they search for a word processor in Software, does LO show up as a
"best bet" or do other applications like abiword or calligra show up? -- a default here may help the user because they are getting the software that may be considered "the best"
+1
Let's also talk about our target users (and run the test Ryan suggested above on them.) We're looking at targeting app developers, right?
- Are app developers typically bandwidth-constrained?
- Do app developers need an office suite? Do they create content using one? Do they consume content that requires having one? (Say a requirements doc from a product manager?)
- If they need it, they have to download it at some point. Either before install, or after install. Is the payload the same whether or not it ships in the image or if it's pull down via yum, right? So if they need the tool, how does pulling it from the image save them bandwidth? (would keeping it in the image save them bandwidth since if they obtained the image via local means / repositories / etc typically available to developers they'd only use internal network and not have to go external?)
- Would an app developer prefer to have the software included in the install image or would they prefer to download it when it was needed?
~m
On Wed, Sep 17, 2014 at 5:08 PM, Máirín Duffy duffy@fedoraproject.org wrote:
Hi,
On 09/16/2014 12:18 PM, Ryan Lerch wrote:
The ease of finding and installing an office suite when there isn't one installed by default would be something that would make a great user test scenario before yanking libreoffice.
Both the scenarios of opening an exisiting file, and trying to create a
document -- some of the things to consider could be:
- what people search for -- (do all users use "Word Processor", or just
"Word")
- if they search for a word processor in Software, does LO show up as a
"best bet" or do other applications like abiword or calligra show up? -- a default here may help the user because they are getting the software that may be considered "the best"
Searching for "word" lists mostly unrelevant stuff at the top of the list, but that's an easy fix - LibreOffice just needs to add "word" as a keyword in the desktop file or appdata.
+1
Let's also talk about our target users (and run the test Ryan suggested above on them.) We're looking at targeting app developers, right?
- Are app developers typically bandwidth-constrained?
No. But targeting developers does not mean we need to provide subpar experience to non-developers. If we want to encourage more people (even people who are not professional developers right now) to become developers, we need to create a platform accessible for everyone - the first step to developing is to be able to use the platform properly.
- Do app developers need an office suite? Do they create content using
one? Do they consume content that requires having one? (Say a requirements doc from a product manager?)
Some do, some don't. I get requirement docs on email or intranet sites,
and sometimes on PDFs, so I don't need an office suite.
- If they need it, they have to download it at some point. Either before
install, or after install. Is the payload the same whether or not it ships in the image or if it's pull down via yum, right? So if they need the tool, how does pulling it from the image save them bandwidth? (would keeping it in the image save them bandwidth since if they obtained the image via local means / repositories / etc typically available to developers they'd only use internal network and not have to go external?)
No, there shouldn't be much difference.
- Would an app developer prefer to have the software included in the
install image or would they prefer to download it when it was needed?
I assume this would vary from person to person. Not all developers are the same person. People testing their apps on VMs will have it easier if our default install would be smaller, for example, because they'll need to allocate less disk space for the VM.
~m
-- desktop mailing list desktop@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/desktop
However, design aside, I assume part of the issue of the immense size of LibreOffice are deps that make little to no sense: Why would an office suite written in C++ and Java need PyXB, which is, according to the package description, "Python package that generates Python source code for classes that correspond to data structures defined by XMLSchema."? This dependency is huge (98MB) and comes with a lot of documentation that should probably be split into a subpackage. I really don't understand why LibreOffice would need that *on runtime*.
On Wed, Sep 17, 2014 at 4:29 PM, Elad Alfassa elad@fedoraproject.org wrote:
On Wed, Sep 17, 2014 at 5:08 PM, Máirín Duffy duffy@fedoraproject.org wrote:
Hi,
On 09/16/2014 12:18 PM, Ryan Lerch wrote:
The ease of finding and installing an office suite when there isn't one installed by default would be something that would make a great user test scenario before yanking libreoffice.
Both the scenarios of opening an exisiting file, and trying to create a document -- some of the things to consider could be:
- what people search for -- (do all users use "Word Processor", or just
"Word")
- if they search for a word processor in Software, does LO show up as a
"best bet" or do other applications like abiword or calligra show up? -- a default here may help the user because they are getting the software that may be considered "the best"
Searching for "word" lists mostly unrelevant stuff at the top of the list, but that's an easy fix - LibreOffice just needs to add "word" as a keyword in the desktop file or appdata.
+1
Let's also talk about our target users (and run the test Ryan suggested above on them.) We're looking at targeting app developers, right?
- Are app developers typically bandwidth-constrained?
No. But targeting developers does not mean we need to provide subpar experience to non-developers. If we want to encourage more people (even people who are not professional developers right now) to become developers, we need to create a platform accessible for everyone - the first step to developing is to be able to use the platform properly.
- Do app developers need an office suite? Do they create content using
one? Do they consume content that requires having one? (Say a requirements doc from a product manager?)
Some do, some don't. I get requirement docs on email or intranet sites, and sometimes on PDFs, so I don't need an office suite.
- If they need it, they have to download it at some point. Either before
install, or after install. Is the payload the same whether or not it ships in the image or if it's pull down via yum, right? So if they need the tool, how does pulling it from the image save them bandwidth? (would keeping it in the image save them bandwidth since if they obtained the image via local means / repositories / etc typically available to developers they'd only use internal network and not have to go external?)
No, there shouldn't be much difference.
- Would an app developer prefer to have the software included in the
install image or would they prefer to download it when it was needed?
I assume this would vary from person to person. Not all developers are the same person. People testing their apps on VMs will have it easier if our default install would be smaller, for example, because they'll need to allocate less disk space for the VM.
~m
-- desktop mailing list desktop@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/desktop
However, design aside, I assume part of the issue of the immense size of LibreOffice are deps that make little to no sense: Why would an office suite written in C++ and Java need PyXB, which is, according to the package description, "Python package that generates Python source code for classes that correspond to data structures defined by XMLSchema."?
I don't have this package installed here. (F20 didn't check rawhide). Also your size calculation's are wrong. Images are compressed so 98MB uncompressed does not mean the image grows by 98MB (stuff like documentation compresses well).
On 09/17/2014 10:29 AM, Elad Alfassa wrote:
On Wed, Sep 17, 2014 at 5:08 PM, Máirín Duffy <duffy@fedoraproject.org mailto:duffy@fedoraproject.org> wrote:
Hi, On 09/16/2014 12:18 PM, Ryan Lerch wrote: The ease of finding and installing an office suite when there isn't one installed by default would be something that would make a great user test scenario before yanking libreoffice. Both the scenarios of opening an exisiting file, and trying to create a document -- some of the things to consider could be: * what people search for -- (do all users use "Word Processor", or just "Word") * if they search for a word processor in Software, does LO show up as a "best bet" or do other applications like abiword or calligra show up? -- a default here may help the user because they are getting the software that may be considered "the best"
Searching for "word" lists mostly unrelevant stuff at the top of the list, but that's an easy fix - LibreOffice just needs to add "word" as a keyword in the desktop file or appdata.
This is why I suggested doing user tests to try to figure out some of these kinks (if we remove LO from the default). Testing will help to show us how people attempt to solve the issue of not having a office suite installed by default.
regards, ryanlerch
On Wed, 2014-09-17 at 11:36 -0400, Ryan Lerch wrote:
This is why I suggested doing user tests to try to figure out some of these kinks (if we remove LO from the default). Testing will help to show us how people attempt to solve the issue of not having a office suite installed by default.
Of course, the easiest thing (at least in the short term) is to just not create that issue in the first place, and keep libreoffice in the default install.
I think we should probably make 'target size' a topic in the next meeting and come to an agreed on number for both download size and install size. Otherwise, we'll keep dithering between 'lets add this useful thing!' and 'lets remove this big thing!'
On Wed, Sep 17, 2014 at 12:19 PM, Matthias Clasen mclasen@redhat.com wrote:
On Wed, 2014-09-17 at 11:36 -0400, Ryan Lerch wrote:
This is why I suggested doing user tests to try to figure out some of these kinks (if we remove LO from the default). Testing will help to show us how people attempt to solve the issue of not having a office suite installed by default.
Of course, the easiest thing (at least in the short term) is to just not create that issue in the first place, and keep libreoffice in the default install.
I think we should probably make 'target size' a topic in the next meeting and come to an agreed on number for both download size and install size. Otherwise, we'll keep dithering between 'lets add this useful thing!' and 'lets remove this big thing!'
Good idea. Added to the list.
josh
On Wed, 2014-09-17 at 12:19 -0400, Matthias Clasen wrote:
I think we should probably make 'target size' a topic in the next meeting and come to an agreed on number for both download size and install size. Otherwise, we'll keep dithering between 'lets add this useful thing!' and 'lets remove this big thing!'
FWIW it should be possible to install writer, calc, impress without installing the database "base" piece, which is the one that sucks in the vast java stack via the pentaho-reporting-flow-engine
C.
On 18 September 2014 09:13, Caolán McNamara caolanm@redhat.com wrote:
FWIW it should be possible to install writer, calc, impress without installing the database "base" piece, which is the one that sucks in the vast java stack via the pentaho-reporting-flow-engine
That's useful to know; I'm never going to use Base. Having uninstalled Base is there an easy way to remove the vast Java stack? I just learned about yum autoremove, but that doesn't seem to have done the job so presumably there are more packages I need to manually remove first...?
Cheers, R
On Thu, 2014-09-18 at 09:13 +0100, Caolán McNamara wrote:
FWIW it should be possible to install writer, calc, impress without installing the database "base" piece, which is the one that sucks in the vast java stack via the pentaho-reporting-flow-engine
We already do not install Base. It's not a tool that's useful or interesting to most users. I think it showed up in F20 during the update with the dep change that pulled in chainsaw etc. (which was reverted long ago).
On Thu, 2014-09-18 at 09:17 -0500, Michael Catanzaro wrote:
On Thu, 2014-09-18 at 09:13 +0100, Caolán McNamara wrote:
FWIW it should be possible to install writer, calc, impress without installing the database "base" piece, which is the one that sucks in the vast java stack via the pentaho-reporting-flow-engine
We already do not install Base. It's not a tool that's useful or interesting to most users. I think it showed up in F20 during the update with the dep change that pulled in chainsaw etc. (which was reverted long ago).
So, why was there so much discussion about java dependencies earlier in this thread ? Are we stilling pulling in java stuff through some other route right now in F21?
C.
On 09/18/2014 05:30 PM, Caolán McNamara wrote:
So, why was there so much discussion about java dependencies earlier in this thread ? Are we stilling pulling in java stuff through some other route right now in F21?
libreoffice-core has a dep on java-headless and libreoffice-ure has a dep on jre-headless, which is what are currently pulling in java.
On Thu, 2014-09-18 at 17:37 +0200, Kalev Lember wrote:
On 09/18/2014 05:30 PM, Caolán McNamara wrote:
So, why was there so much discussion about java dependencies earlier in this thread ? Are we stilling pulling in java stuff through some other route right now in F21?
libreoffice-core has a dep on java-headless and libreoffice-ure has a dep on jre-headless, which is what are currently pulling in java.
Which is the minimal java stack, the original post said "its various java dependencies" which made me assume it was the full base stack. So... is LibreOffice the only thing pulling in jre-headless in the default install ?
Because we could probably get rid of those requires if we need to. We removed and converted a lot of stuff away from java since the time the original requires were added, though it is convenient from the perspective of supporting manually added extensions to know that java is installed. I guess we're already compromised in the sense that we now only know for sure that at least the headless variant is installed.
C.
Hi
On Thu, Sep 18, 2014 at 12:32 PM, Caolán McNamara wrote:
Because we could probably get rid of those requires if we need to. We removed and converted a lot of stuff away from java since the time the original requires were added, though it is convenient from the perspective of supporting manually added extensions to know that java is installed. I guess we're already compromised in the sense that we now only know for sure that at least the headless variant is installed.
Yes, I think dropping the requires is a useful thing. A release note must be added as a warning for users installing third party extensions that may still rely on Java
Rahul
On Fri, Sep 19, 2014 at 7:18 AM, Rahul Sundaram metherid@gmail.com wrote:
Hi
On Thu, Sep 18, 2014 at 12:32 PM, Caolán McNamara wrote:
Because we could probably get rid of those requires if we need to. We removed and converted a lot of stuff away from java since the time the original requires were added, though it is convenient from the perspective of supporting manually added extensions to know that java is installed. I guess we're already compromised in the sense that we now only know for sure that at least the headless variant is installed.
Yes, I think dropping the requires is a useful thing. A release note must be added as a warning for users installing third party extensions that may still rely on Java
Well when someone installs a third party extension and have no java would libreoffice tell him to install java? If not things will just fail without an obvious reason making our product looks bad. So I disagree on it being a useful thing. We don't have to save every single megabyte no matter what. And no a release notes entry is not enough.
On 09/19/2014 07:50 AM, drago01 wrote:
On Fri, Sep 19, 2014 at 7:18 AM, Rahul Sundaram metherid@gmail.com wrote:
On Thu, Sep 18, 2014 at 12:32 PM, Caolán McNamara wrote:
Because we could probably get rid of those requires if we need to. We removed and converted a lot of stuff away from java since the time the original requires were added, though it is convenient from the perspective of supporting manually added extensions to know that java is installed. I guess we're already compromised in the sense that we now only know for sure that at least the headless variant is installed.
Yes, I think dropping the requires is a useful thing. A release note must be added as a warning for users installing third party extensions that may still rely on Java
Well when someone installs a third party extension and have no java would libreoffice tell him to install java?
It depends. Typically, extensions would install just fine without any notice (old extensions using active registration could instead cause a rather cryptic error message and just not install at all). They would only complain once their functionality is actually used. Depending on how that functionality would be triggered, they'd either pop up a nice dialog stating that no JRE is available, or e.g. in the -- probably common -- case of triggering such functionality via a menu entry would just silently do nothing at all. (And with the recent move to a java-headless--only dependency, extensions bringing along their own Java-drawn GUI will fail as badly or as gracefully as they handle a ClassNotFoundError.)
So, it's rather the "just fail without any obvious reason" case for now, I'd say.
If not things will just fail without an obvious reason making our product looks bad. So I disagree on it being a useful thing. We don't have to save every single megabyte no matter what. And no a release notes entry is not enough.
Stephan
On Fri, 2014-09-19 at 09:38 +0200, Stephan Bergmann wrote:
And with the recent move to a java-headless--only dependency, extensions bringing along their own Java-drawn GUI will fail as badly or as gracefully as they handle a ClassNotFoundError.
Ideally some package-kit magic would kick in "somehow" to prompt to install the missing bit. But no idea how that would work.
C.
So those java packages we're discussing. How much space do they take anyway?
It seems to me that having third party extensions silently failing to save a dozen megs might not be worth it. If we're talking about 200megs on the other hand...
On Fri, 2014-09-19 at 09:14 +0100, Caolán McNamara wrote:
On Fri, 2014-09-19 at 09:38 +0200, Stephan Bergmann wrote:
And with the recent move to a java-headless--only dependency, extensions bringing along their own Java-drawn GUI will fail as badly or as gracefully as they handle a ClassNotFoundError.
Ideally some package-kit magic would kick in "somehow" to prompt to install the missing bit. But no idea how that would work.
C.
On Fri, Sep 19, 2014 at 1:57 PM, Alberto Ruiz aruiz@redhat.com wrote:
So those java packages we're discussing. How much space do they take anyway?
It seems to me that having third party extensions silently failing to save a dozen megs might not be worth it. If we're talking about 200megs on the other hand...
still not worth it even if it where 200MB ... either it is required or it isn't. We shouldn't cripple the user experience just to save disk space. 200MB is not even worth talking about much in 2014.
On Fri, Sep 19, 2014 at 5:27 PM, Alberto Ruiz aruiz@redhat.com wrote:
It seems to me that having third party extensions silently failing to save a dozen megs might not be worth it. If we're talking about 200megs on the other hand...
What extensions do we ship with LibreOffice that requires Java? If they are third-party and upstream doesn't compel to include them by default do we really need them.
On Arch Linux, third party extensions rely on whatever Java installed by the user. There are products like Android Studio that don't support OpenJDK and in such scenario you end up installing both Java but using only one. That leads to maintenance overhead.
Hello Sudhir,
I am afraid that you're missing one important point.
If a user installs a 3rd party plugin that is expected to work and it doesn't work that's bad. From her/his POV Fedora would be giving her/him a worse experience than other distros or OSes where these plugins work out of the box in LibreOffice.
A good default experience is certainly more valuable than saving a few megs. (Hence my question, how much space are we talking about here, as I suspect we're trying to overoptimize).
On Fri, 2014-09-19 at 17:52 +0530, Sudhir Khanger wrote:
On Fri, Sep 19, 2014 at 5:27 PM, Alberto Ruiz aruiz@redhat.com wrote:
It seems to me that having third party extensions silently failing to save a dozen megs might not be worth it. If we're talking about 200megs on the other hand...
What extensions do we ship with LibreOffice that requires Java? If they are third-party and upstream doesn't compel to include them by default do we really need them.
On Arch Linux, third party extensions rely on whatever Java installed by the user. There are products like Android Studio that don't support OpenJDK and in such scenario you end up installing both Java but using only one. That leads to maintenance overhead.
-- Regards, Sudhir Khanger. sudhirkhanger.com https://github.com/donniezazen
On Fri, Sep 19, 2014 at 6:44 PM, Alberto Ruiz aruiz@redhat.com wrote:
If a user installs a 3rd party plugin that is expected to work and it doesn't work that's bad. From her/his POV Fedora would be giving her/him a worse experience than other distros or OSes where these plugins work out of the box in LibreOffice.
Those third party packages should have Java as a dependency. They should check for 'java-environment' and not explicitly OpenJDK.
On Fri, Sep 19, 2014 at 5:03 PM, Sudhir Khanger sudhir@sudhirkhanger.com wrote:
On Fri, Sep 19, 2014 at 6:44 PM, Alberto Ruiz aruiz@redhat.com wrote:
If a user installs a 3rd party plugin that is expected to work and it doesn't work that's bad. From her/his POV Fedora would be giving her/him a worse experience than other distros or OSes where these plugins work out of the box in LibreOffice.
Those third party packages should have Java as a dependency. They should check for 'java-environment' and not explicitly OpenJDK.
Those aren't rpms.
On 19 September 2014 13:22, Sudhir Khanger sudhir@sudhirkhanger.com wrote:
On Fri, Sep 19, 2014 at 5:27 PM, Alberto Ruiz aruiz@redhat.com wrote: There are products like Android Studio that don't support OpenJDK and in such scenario you end up installing both Java but using only one. That leads to maintenance overhead.
Indeed, JetBrains IntelliJ doesn't support OpenJDK either, so with LibreOffice installed one ends-up with two versions of Java required, and having to go through the hassle of configuring alternatives. It would be nice if there was some mechanism to use Oracle's JDK to satisfy package dependencies if one has chosen to install it so that OpenJDK could be removed.
Oracle's JDK (the full JDK - not just the JRE) weighs in at just shy of 300Mb.
On Fri, Sep 19, 2014 at 3:17 PM, Richard Turner rjt@zygous.co.uk wrote:
On 19 September 2014 13:22, Sudhir Khanger sudhir@sudhirkhanger.com wrote:
On Fri, Sep 19, 2014 at 5:27 PM, Alberto Ruiz aruiz@redhat.com wrote: There are products like Android Studio that don't support OpenJDK and in such scenario you end up installing both Java but using only one. That leads to maintenance overhead.
Indeed, JetBrains IntelliJ doesn't support OpenJDK [...]
What does "doesn't support" mean? Does it not run at all? (whats missing?) Is something broken? Or do they just not care about bug reports when someone uses openjdk but it otherwise works?
On 19 September 2014 14:23, drago01 drago01@gmail.com wrote:
On Fri, Sep 19, 2014 at 3:17 PM, Richard Turner rjt@zygous.co.uk wrote:
What does "doesn't support" mean? Does it not run at all? (whats missing?) Is something broken? Or do they just not care about bug reports when someone uses openjdk but it otherwise works?
My experience has been that they're quite closed-lipped about OpenJDK. They don't care about bug reports, no, but they also state that there are 'known issues' when running under OpenJDK. I tried just today and it does run, but I was finding that it would hang fairly frequently and I'd have to kill -s 9 it.
To be fair, stability using Oracle's JDK isn't great at the moment: ibus doesn't seem to want to play nicely, which doesn't hang the app, but does prevent keyboard input. Restarting ibus then crashes the app. I don't know whether this is actually an ibus problem though, and the OpenJDK hangs vs. the Oracle JDK crashes are symptoms of the same problem being handled differently.
On Fri, Sep 19, 2014 at 6:53 PM, drago01 drago01@gmail.com wrote:
What does "doesn't support" mean? Does it not run at all? (whats missing?) Is something broken? Or do they just not care about bug reports when someone uses openjdk but it otherwise works?
They will not take bug reports. Works or doesn't work depends on your mileage.
On 09/19/2014 02:22 PM, Sudhir Khanger wrote:
On Fri, Sep 19, 2014 at 5:27 PM, Alberto Ruiz aruiz@redhat.com wrote:
It seems to me that having third party extensions silently failing to save a dozen megs might not be worth it. If we're talking about 200megs on the other hand...
What extensions do we ship with LibreOffice that requires Java? If they are third-party and upstream doesn't compel to include them by default do we really need them.
Third party LibreOffice extensions are something that upstream by design does not include---they e.g. cover functionality that is deemed too specialized to warrant general inclusion. And we as Fedora don't ship them either---they're an area where repackaging as RPMs probably doesn't make sense, at least not exclusively.
(Related to the question of how much space would be saved by dropping java-headless, LO's move from depending java to java-headless wasn't so much motivated by space savings but rather to work around "Java package [...] launchers we don't want to have in our default install," https://bugzilla.redhat.com/show_bug.cgi?id=1131425.)
(And, just to be clear, LO works with whatever JRE's available at runtime, OpenJDK, Oracle, even GCJ still.)
Stephan
On 09/17/2014 10:29 AM, Elad Alfassa wrote:
Searching for "word" lists mostly unrelevant stuff at the top of the list, but that's an easy fix - LibreOffice just needs to add "word" as a keyword in the desktop file or appdata.
Ok. Fixing this should probably be a prerequisite for dropping LO.
No. But targeting developers does not mean we need to provide subpar experience to non-developers. If we want to encourage more people (even people who are not professional developers right now) to become developers, we need to create a platform accessible for everyone - the first step to developing is to be able to use the platform properly.
If you want to support non-developers / less technical folks who are heavy office document users, removing LO from the default install is a no-go. I support these folks using LO as part of my job. They get it pre-installed on their RHEL CSBs and I am really not sure they'd be super comfortable installing it on their own if they tried out Fedora at home. If we are concerned about this audience we should definitely run a few (quick, easy, doesn't have to be a huge investment) tests on them to see what the experience without LO installed by default is for them and how they cope so we know what we have to do to make it easier.
- Do app developers need an office suite? Do they create content using one? Do they consume content that requires having one? (Say a requirements doc from a product manager?)
Some do, some don't. I get requirement docs on email or intranet sites, and sometimes on PDFs, so I don't need an office suite.
I'm not a developer but involved in the development process and I have definitely gotten .doc files from customers and PMs (reqs documents), slide decks from developers for conference presos, etc. But I don't know how universal this is. Maybe worth doing a quick survey?
- If they need it, they have to download it at some point. Either before install, or after install. Is the payload the same whether or not it ships in the image or if it's pull down via yum, right? So if they need the tool, how does pulling it from the image save them bandwidth? (would keeping it in the image save them bandwidth since if they obtained the image via local means / repositories / etc typically available to developers they'd only use internal network and not have to go external?)
No, there shouldn't be much difference.
- Would an app developer prefer to have the software included in the install image or would they prefer to download it when it was needed?
I assume this would vary from person to person. Not all developers are the same person. People testing their apps on VMs will have it easier if our default install would be smaller, for example, because they'll need to allocate less disk space for the VM.
Oh geez thats a really good point. But are the devs going to use the workstation in a vm? Or are they going to use server or cloud in a vm? I mean, the majority of devs we're targeting are web app devs right? Not Linux desktop devs... so they'd be doing cloud/server VMs I would think?
~m
On Wed, Sep 17, 2014 at 8:26 PM, Máirín Duffy duffy@fedoraproject.org wrote:
[..]
I assume this would vary from person to person. Not all developers are the same person. People testing their apps on VMs will have it easier if our default install would be smaller, for example, because they'll need to allocate less disk space for the VM.
Oh geez thats a really good point. But are the devs going to use the workstation in a vm? Or are they going to use server or cloud in a vm?
Neither ... most would just use light weight alternatives like docker containers. VMs are only useful if you are developing for a different platform (like Windows) or you are developing the OS itself (and OS developers are not the type of developers we target).
If we don't install libreoffice by default we'd save around ~500MB (of libreoffice itself and it's various java dependencies) from the live media. This means less strain on mirrors, and that people will get their media faster (even if you have a really fast connection - smaller downloads finish sooner).
At the very least it should be made possible to opt out of Libreoffice when doing a Workstation netinstall (it is for a Gnome Desktop install).
Jens
On Sep 17, 2014, at 4:18 AM, Jens Petersen petersen@redhat.com wrote:
If we don't install libreoffice by default we'd save around ~500MB (of libreoffice itself and it's various java dependencies) from the live media. This means less strain on mirrors, and that people will get their media faster (even if you have a really fast connection - smaller downloads finish sooner).
At the very least it should be made possible to opt out of Libreoffice when doing a Workstation netinstall (it is for a Gnome Desktop install).
The netinst media installer main menu has a "Software Selection" option, and in there you can uncheck LibreOffice under Add-Ons. The Live media installer doesn't offer this.
Chris Murphy
On Sep 17, 2014, at 4:18 AM, Jens Petersen petersen@redhat.com wrote:
At the very least it should be made possible to opt out of Libreoffice when doing a Workstation netinstall (it is for a Gnome Desktop install).
The netinst media installer main menu has a "Software Selection" option, and in there you can uncheck LibreOffice under Add-Ons. The Live media installer doesn't offer this.
Chris: the optional groups available vary with the different environment groups.
Anyway I went ahead and moved libreoffice to be an default optional group for Workstation:
https://git.fedorahosted.org/cgit/comps.git/commit/?id=a34183c0c20696d4c0d6b...
so hopefully this should work now.
Jens
ps Do we actually still want to keep the GNOME Desktop environment group around (in addition to the new Workstation environment)? Or do we leave it until the F22 cycle? I don't really mind: I guess the main focus is on Live anyway.
On 09/18/2014 03:35 AM, Jens Petersen wrote:
Anyway I went ahead and moved libreoffice to be an default optional group for Workstation:
https://git.fedorahosted.org/cgit/comps.git/commit/?id=a34183c0c20696d4c0d6b...
I am not sure bringing back package selection [1] in Anaconda is a step in the right direction. I'd rather have all customization done post-install through GNOME Software and keep the first installation experience as straightforward as possible.
In fact, I would like Anaconda to completely hide the software selection spoke when there is nothing optional to choose, as is (was?) the case with Workstation.
On 09/18/2014 05:27 AM, Kalev Lember wrote:
I am not sure bringing back package selection [1] in Anaconda is a step in the right direction. I'd rather have all customization done post-install through GNOME Software and keep the first installation experience as straightforward as possible.
In fact, I would like Anaconda to completely hide the software selection spoke when there is nothing optional to choose, as is (was?) the case with Workstation.
We (the team working on the Anaconda UI rewrite) explicitly excluded package-level selection for UX reasons (the way RPM & depsolving works it basically lies to users and can never meet the expectations the UI necessarily has to set) and I think bringing it back is a no-go. If we wanted to build some kind of tool for tweakers that lets you do package level selection and then builds an image, I think that would be the right way to solve that issue. Anaconda is too limited an environment for users to have the information they need to make informed selections and it's also too limited an environment to properly set expectations about the dep-based nature of RPM and how what you select and de-select in the UI doesn't mean the thing you de-selected isn't going to be present....
Would love for the software selection spoke to be completely hidden when there's nothing to choose. We hide it in live install images IIRC.
~m
Hi Kalev,
I am not sure bringing back package selection [1] in Anaconda is a step in the right direction. I'd rather have all customization done post-install through GNOME Software and keep the first installation experience as straightforward as possible.
I think it is nice to keep optional groups visible for netinstall: that is often why people choose to use it rather than the Live image. In fact I would not mind seeing KDE listed there too for example.
In fact, I would like Anaconda to completely hide the software selection spoke when there is nothing optional to choose, as is (was?) the case with Workstation.
It was already there.
Jens
desktop@lists.fedoraproject.org