[Standards] Re: XEP-0073: Question about service discovery

JD Conley jd.conley at coversant.net
Thu Feb 8 19:23:31 UTC 2007

> > I realize I'm probably on the verge of getting pelted with produce
> > for mentioning this once more, but I'm going to point out -- again
> > -- that a yearly certification process might solve this.  In 2007,
> > one could make PEP a 'recommended' feature for servers, and PEP
> > avatars a 'recommended' feature for clients.  Then in 2008, one
> > could have it be a requirement; anyone who didn't implement it in
> > 2007 doesn't get certified in 2008.  In 2009, you could even make a
> > recommendation that clients /stop/ using vcard-temp hashes or they
> > lose their certification.
> "...everyone who didn't implement it in 2007 either has to do the
> work now, or doesn't get certified in 2008..."
> A few important words missing there, but I think my meaning is
> probably taken. ;)

Rachel -- I'll take some of that produce too. My 2c...

Those of us whose livelihood depends on this stuff do not really have
the time to circle back on all these features that already work. It's
hard for users/investors/me to fathom "uhh, yeah, we spent a month
replacing old avatar code so you can have the same features but we use
less bandwidth sometimes and it's much sexier! -- oh and it doesn't work
with all those other clients/servers out there anymore." The end user
and business value just isn't there. Now, if it's a new protocol suite
that offers good value (MUC vs. groupchat, pubsub vs. public
storage/presence, jingle vs. s5b/etc), then those are things that will
get implemented more quickly. It's a gradual, evolutionary process.

Don't get me wrong. I love the PEP concept. There are some really neat
things you can do with it, but you're not going to see people ripping
out vcard-temp and x:avatar code for the sexy PEP version until years
after all the legacy stuff is not in use in the wild. Heck, we still get
people asking about installing our software on Win9x and NT4...
If you build a product using only the latest and greatest protocols
(like PEP), with no backward compatibility, more power to ya.

A certification process is definitely still in order. We have hundreds
of specs out there. The tech tree is very complex. To maintain
interoperability in the future I think the cert process is a must.


More information about the Standards mailing list