[ClusterLabs] Antw: [EXT] Re: Q: All the versions

[Ken Gaillot]

On Thu, 2020-03-19 at 15:26 +0100, Ulrich Windl wrote:
> > > > Ken Gaillot <kgaillot at redhat.com> schrieb am 18.03.2020 um
> > > > 17:59 in
> 
> Nachricht
> <5851_1584550779_5E72537B_5851_802_1_3128245adeaa62fd81c983c301a5acf3
> 6e59ff62.ca
> el at redhat.com>:
> > On Wed, 2020‑03‑18 at 10:31 +0100, Ulrich Windl wrote:
> > > Hi!
> > > 
> > > I have a question on the various versions in the RA metadata. For
> > > example:
> > > 
> > > <?xml version="1.0"?>
> > > <!DOCTYPE resource‑agent SYSTEM "ra‑api‑1.dtd">
> > > <resource‑agent name="HealthCPU" version="0.1">
> > > <version>1.0</version>
> > > 
> > > The XML version should be OK. Next version is the RA API version
> > > 1 as
> > > indicated in the file name.
> > > Then the resource‑agent element has version "0.1".
> > > Is that the version (0.1) that should be incremented if the
> > > metadata
> > > changes or the contents of the <version> element (1.0)?
> > > 
> > > Confused, sorry!
> > > 
> > > Regards,
> > > Ulrich
> > 
> > It is definitely confusing.
> > 
> > The version inside <resource‑agent> is for the agent's own use, so
> > the
> > agent developer can do whatever they want with that one.
> > 
> > The <version> tag is for the RA API version (it happens to be in
> > the
> > DTD filename as well, but what that file is named is not considered
> > to
> > have any significance). If possible I'd rename this to be clearer
> > but
> > that's probably not a good idea due to backward compatibility
> > issues.
> 
> But that's against the tradition of SGML/XML usage: If there are
> different
> DTDs, then the doctype expresses that, probably using a "public"
> identifier
> instead of a "system" identifier. An extra <version> element inside
> the
> document is redundant at least...
>
> Regards,> Ulrich

Now that I think about it, they are different. The dtd version is the
version of the API's meta-data XML syntax, whereas <version> is the
version of the entire API including e.g. what actions an agent must
support, where they must be located, etc.
-- 
Ken Gaillot <kgaillot at redhat.com>