Hi,
Jindrich Novy has been preparing texlive packages for F8 for a while now, and there are essentially 3 RPMs:
telive (the binaries) texlive-texmf (the texmf tree) and texlive-texmf-errata
The one I am concerned about is telive-texmf-errata. As Jindrich says "This is the errata package for the TeXLive 2007 formatting system. The purpose of this package is to support updates to huge texmf tree without a need to download all the texmf tree again, but to ship only the fixed parts. texlive-errata puts updated files into a seperate texmf tree which is searched prior to the main tree so there are no conflicts between texlive-errata and texlive-texmf packages."
I think this is a totally different packaging paradigm - as far as I'm aware there's no precedent in Fedora for issuing errata packages rather than updated packages. A far better alternative IMO is to have finer grained subpackaging of the texlive texmf tree, such that updates don't replace the whole thing. That of course has other major advantages, such as allowing smaller tex installs.
Also, to have *two* system managed texmf trees searched is a big change, and something else that system admins have to think about when they add their own local texmf trees.
Put more bluntly, while I understand the convenience from a packagers point of view, this seems like a really ugly way to package. It feels a bit like the ever increasing number of hotfixes you get installed on an M$ system (although there would never be more than a single texmf-errata package installed at a time of course).
I know Rex Dieter likes the errata package idea. I wonder what others think?
Cheers, Jonathan
On Mon, Aug 06, 2007 at 11:28:38PM +0100, Jonathan Underwood wrote:
Hi,
I know Rex Dieter likes the errata package idea. I wonder what others think?
I think that such decisions should be left to the packagers, as long as it is not obviously wrong. A bit like split choices.
-- Pat
On Tue, 2007-08-07 at 00:31 +0200, Patrice Dumas wrote:
On Mon, Aug 06, 2007 at 11:28:38PM +0100, Jonathan Underwood wrote:
I know Rex Dieter likes the errata package idea. I wonder what others think?
I think that such decisions should be left to the packagers, as long as it is not obviously wrong. A bit like split choices.
Hey, OOo dictionaries are big... let's make errata packages for them differently for updates. Maybe for the data for $game, too.
...
I think that this is a pretty bad idea for us to follow down. Much like we package perl modules natively rather than telling people to use CPAN, we should be handling updates to packages natively rather than errata packages that stand along-side. If the argument is size and space, then help out with testing presto and getting the support into the buildsystem so that we can have it enabled by default and helping for *all* packages rather than just a select few that have built their own way of doing things
Jeremy
On Mon, Aug 06, 2007 at 08:54:45PM -0400, Jeremy Katz wrote:
On Tue, 2007-08-07 at 00:31 +0200, Patrice Dumas wrote:
On Mon, Aug 06, 2007 at 11:28:38PM +0100, Jonathan Underwood wrote:
I know Rex Dieter likes the errata package idea. I wonder what others think?
I think that such decisions should be left to the packagers, as long as it is not obviously wrong. A bit like split choices.
Hey, OOo dictionaries are big... let's make errata packages for them differently for updates. Maybe for the data for $game, too.
If it makes sense for a specific package, yes.
...
I think that this is a pretty bad idea for us to follow down. Much like we package perl modules natively rather than telling people to use CPAN,
That's a different issue, still use rpm here.
we should be handling updates to packages natively rather than errata packages that stand along-side.
In genenal, yes, but leave it to packagers when they feel strongly about it. For texlive it may make sense.
If the argument is size and space, then help out with testing presto and getting the support into the buildsystem so that we can have it enabled by default and helping for *all* packages rather than just a select few that have built their own way of doing things
I am not sure that using presto is the answer here. Doing a texlive errata package solves more than the space issue. Moreover at any point the errata may be integrated in the main rpm.
I don't know texlive a lot, but, in the general case using a trick like an errata package may help updating only part of the package when upstream releases errata and keeps a monolithic package otherwise.
-- Pat
On Mon, 2007-08-06 at 20:54 -0400, Jeremy Katz wrote:
On Tue, 2007-08-07 at 00:31 +0200, Patrice Dumas wrote:
On Mon, Aug 06, 2007 at 11:28:38PM +0100, Jonathan Underwood wrote:
I know Rex Dieter likes the errata package idea. I wonder what others think?
I think that such decisions should be left to the packagers, as long as it is not obviously wrong. A bit like split choices.
Hey, OOo dictionaries are big... let's make errata packages for them differently for updates. Maybe for the data for $game, too.
...
I think that this is a pretty bad idea for us to follow down. Much like we package perl modules natively rather than telling people to use CPAN, we should be handling updates to packages natively rather than errata packages that stand along-side. If the argument is size and space, then help out with testing presto and getting the support into the buildsystem so that we can have it enabled by default and helping for *all* packages rather than just a select few that have built their own way of doing thing
Another alternative is to package the texmf tree more modular, which would not only help to reduce space consumption for updates, but also for the initial installation - not everybody who uses latex also has a burning need for context...
On Mon, Aug 06, 2007 at 09:28:53PM -0400, Matthias Clasen wrote:
Another alternative is to package the texmf tree more modular, which would not only help to reduce space consumption for updates, but also for the initial installation - not everybody who uses latex also has a burning need for context...
+1
Further modularization seems to be the best way how to do this in the future. The current subpackage structure is designed for easy adoption by the packages dependent on tetex so that it's almost identical. The current strategy is to replace tetex with texlive in F8 in a way that a minimal overhead is needed from the other package maintainers, so that it can be done smoothly and quickly.
Jindrich
On Mon, Aug 06, 2007 at 08:54:45PM -0400, Jeremy Katz wrote:
On Tue, 2007-08-07 at 00:31 +0200, Patrice Dumas wrote:
On Mon, Aug 06, 2007 at 11:28:38PM +0100, Jonathan Underwood wrote:
I know Rex Dieter likes the errata package idea. I wonder what others think?
I think that such decisions should be left to the packagers, as long as it is not obviously wrong. A bit like split choices.
Hey, OOo dictionaries are big... let's make errata packages for them differently for updates. Maybe for the data for $game, too.
...
I think that this is a pretty bad idea for us to follow down. Much like we package perl modules natively rather than telling people to use CPAN, we should be handling updates to packages natively rather than errata packages that stand along-side. If the argument is size and space, then help out with testing presto and getting the support into the buildsystem so that we can have it enabled by default and helping for *all* packages rather than just a select few that have built their own way of doing things
You miss the point here. It's not about introducing a new packaging paradigm, that other packagers should adopt. It's solely for the texlive packaging purpose in the current state of texlive and the buildsystem. Presto is most likely useless here as many files in the noarch packages are configs/ghosted and are expected to be modified by the regular usage, so deltarpm wouldn't help here and it would download the full packages anyway in many cases.
Jindrich
On Mon, Aug 06, 2007 at 08:54:45PM -0400, Jeremy Katz wrote:
On Tue, 2007-08-07 at 00:31 +0200, Patrice Dumas wrote:
On Mon, Aug 06, 2007 at 11:28:38PM +0100, Jonathan Underwood wrote:
I know Rex Dieter likes the errata package idea. I wonder what others think?
I think that such decisions should be left to the packagers, as long as it is not obviously wrong. A bit like split choices.
Hey, OOo dictionaries are big... let's make errata packages for them differently for updates.
Oh yes, please, I have the feeling that all I'm updating lately are these dictionaries ;)
I think for texmf the errata method is OK, after all it's one of the largest packages (maybe the largest after openoffice?), and it would allow the maintainer to be able to follow closer the texlive updates instead of holding back updates because updates punish more than they help.
packaging@lists.fedoraproject.org