----- Original Message -----
From: "Adam Litke" agl@us.ibm.com To: "Saggi Mizrahi" smizrahi@redhat.com Cc: vdsm-devel@lists.fedorahosted.org, "Vinzenz Feenstra" vfeenstr@redhat.com Sent: Tuesday, January 15, 2013 9:27:31 AM Subject: Re: [vdsm] API Documentation & Since tag
On Mon, Jan 14, 2013 at 05:45:45PM -0500, Saggi Mizrahi wrote:
----- Original Message -----
From: "Adam Litke" agl@us.ibm.com To: "Saggi Mizrahi" smizrahi@redhat.com Cc: vdsm-devel@lists.fedorahosted.org, "Vinzenz Feenstra" vfeenstr@redhat.com Sent: Monday, January 14, 2013 5:21:41 PM Subject: Re: [vdsm] API Documentation & Since tag
On Mon, Jan 14, 2013 at 12:37:57PM -0500, Saggi Mizrahi wrote:
----- Original Message -----
From: "Adam Litke" agl@us.ibm.com To: "Vinzenz Feenstra" vfeenstr@redhat.com Cc: vdsm-devel@lists.fedorahosted.org Sent: Friday, January 11, 2013 9:03:19 AM Subject: Re: [vdsm] API Documentation & Since tag
On Fri, Jan 11, 2013 at 10:19:45AM +0100, Vinzenz Feenstra wrote:
Hi everyone,
We are currently documenting the API in vdsmapi-schema.json I noticed that we have there documented when a certain element newly is introduced using the 'Since' tag. However I also noticed that we are not documenting when a field was newly added, nor do we update the 'since' tag.
We should start documenting in what version we've introduced a field. A suggestion by saggi was to add to the comment for example: @since: 4.10.3
What is your point of view on this?
I do think it's a good idea to add this information. How about supporting multiple Since lines in the comment like the following made up example:
## # @FenceNodePowerStatus: # # Indicates the power state of a remote host. # # @on: The remote host is powered on # # @off: The remote host is powered off # # @unknown: The power status is not known # # @sentient: The host is alive and powered by its own metabolism # # Since: 4.10.0 - @FenceNodePowerStatus # Since: 10.2.0 - @sentient ##
I don't like the fact that both lines don't point to the same type of token. I also don't like that it's a repeat of the type names and field names.
I prefer Vinzenz original suggestion (on IRC) of moving the "Since" token up and then have it be a state. It also makes discerning what entities you can use up to a certain version easier if you make sure to keep them sorted.
We can do this because the order of the fields and availability is undetermined (unlike real structs).
That is not correct. These structures are parsed into an OrderedDict and the ordering is important (especially for languages like C which might use real structs).
The "wire" format, json, ignores the ordering, further more, for languages like C we can't use actual structs because then we have to bump a major version every time we add a field as the sizeof(struct Foo) changed
## # @FenceNodePowerStatus: # # Indicates the power state of a remote host. # # Since: 4.10.0 # # @on: The remote host is powered on # # @off: The remote host is powered off # # @unknown: The power status is not known # # Since: 10.2.0 # # @sentient: The host is alive and powered by its own metabolism # ##
The problem though is that it makes since a property of the fields and not of the struct. This isn't that much of a problem as we can assume the earliest version is the time when the struct was introduced.
I don't like this any better than my suggestion. Aside from the fact that field ordering is important (in the data structure itself), this spreads the since information throughout the comment rather than concentrating it in a single place.
Well, thinking about it, I don't understand why structs need to have a "Since" property anyway. Only verbs should have it. Structs are available (by inference) since the earliest call that produces them.
All fields in a struct are optional anyway. Old versions wouldn't try and access them, new clients should always assume these fields may not be returned anyway.
All _newly_added_ fields must be optional. Fields that are part of the original definition of the type can be required fields. This reminds me that we will need to audit the schema for fields that can be made optional. For example, when creating Vm*Device objects, the VmDeviceAddress member can be omitted.
I think the best solution is what python is doing in it's documentation: You have "versionadded" which is the version where the entity has been introduced. Subsequent changes use multiple lines of "versionchanged" and a description.
## # @FenceNodePowerStatus: # # Indicates the power state of a remote host. # # @on: The remote host is powered on # # @off: The remote host is powered off # # @unknown: The power status is not known # # @sentient: The host is alive and powered by its own metabolism # # Since: 4.10.0 # # Changed: 10.4.21 added the @sentient possibility. # ##
I just noticed that the sample code is actually an enum. Enums that are returned by VDSM can't be changed since old clients will not be able to expect\handle them. We either need to create a new enum (as it is in this case) where sentient either maps to @on or @unknown. Or we make @FenceNodePowerStatus a gradient where positive values means @on and negative values mean @off, 0 is @unknown and every value we add is a sub-state of it (@sentient >= 100, @barelyOn = 1). This way old client just see it as @on but newer clients can discern between classic @on and @sentient. This is similar to how log levels work.
This means that in order for an enum to be mutable it has to document how it might change.
It also raises the question about how to represent this in the schema, we also need to decide whether the conversion is done by libvdsm and always return an enum or should we return a number at let the user figure it out.
## # @FenceNodePowerStatus: # # Indicates the power state of a remote host. Positive values mean on and # negative values mean off. 0 means the state is unknown. Values over 100 # are reserved of states where the host is on and is powered by it's own # metabolism. Values under 100 mean that the host is dead. # # @on: The remote host is powered on # # @off: The remote host is powered off # # @unknown: The power status is not known # # @sentient: The host is alive and powered by its own metabolism # # @dead: The host is off and can never be on again. Can only happen if # fence node was previously in @sentient state. # # Since: 4.10.0 # # Changed: 10.4.21 Added the @sentient possibility. Older versions # are unable to return this value. # # Changed: 22.4.21 Added the @dead possibility due users wishing to know when # to morn @sentient fence nodes. Older versions will return # regular @off state and in these cases it's the caller's # responsibility to detect if the agent is @dead or just @off. # ##
Errors are a good example where this is a real problem, if you add a new error old clients can be baffled by it. Taking a note from HTTP errors we need to define error classes and all errors have to be classified.
I have no idea how to represent this in the schema.
-- Adam Litke agl@us.ibm.com IBM Linux Technology Center