[jdev] Re: Re: The State of Our Code-bases

[Lucas Nussbaum]

On Thu, Sep 02, 2004 at 12:48:39AM -0500, Nolan Eakins <sneakin at semanticgap.com> wrote:
> Lucas Nussbaum wrote:
> > 1) Jabber is a difficult technology. It is very difficult for end users
> > to understand what it's all about. I think we lack a good documentation
> > for end users, covering stuff like what is a Jabber server, a JID, can I
> > communicate with users on other servers, etc ... Or I haven't found
> > it yet. If such a documentation exists, it would be interesting to start
> > a translation project, so Jabber-related websites could use localized
> > versions.
> 
> Does email have good documentation? If you're online, you know what email
> is. Why? Because any ISP will give you at least one email address to use,
> and then name off a couple of email programs to use and how to set them up.
> If an ISP gave you a Jabber account to use, presumably it'll be the same as
> your email. So JIDs and what is a Jabber server wouldn't need to be
> documented for the end user.

Email can be understood with real world experience since it has a lot of
similarities with snail mail :
email address <-> postal address
mailbox <-> mailbox
etc
but what about :
jabber server ? transports ?

Currently no ISP gives Jabber accounts. So what we have to deal with is
users who decided to use Jabber, and those users need to grasp far too
many concepts.

> So why aren't ISPs giving out Jabber accounts yet? That's gotta do with the
> state of our code-bases. An ISP isn't going to setup a buggy, incomplete,
> and hard to manage server. They need the equivalent of an Apache, something
> they do setup and use.

I think a poll would be nice here. I think you are one step too far with
this ISP story : ISPs don't see the need for jabber servers since nobody
uses Jabber. And nobody uses Jabber because it is difficult to grasp and
poorly documented.

> > 2) The number of clients/libraries/servers. For somebody coming to
> > Jabber, it is very difficult to choose a client. There's no comparative
> > charts saying "this one does this, but doesn't do this, etc ...". There's
> > http://www.jabber.org/software/clients.php, but I find it quite useless.
> > There's the same problem for libraries. There are for example 4
> > different libraries in python (jabberpy, xmpppy, pyxmpp, twisted). How
> > do they compare with each other ?
> 
> Tell your friends I use this client. Set it up for them. Get them to use it.
> If they like it, they won't go looking for a new one. If they do go
> looking, maybe they'll find something better or worse. Word of mouth should
> be good, except for the isolated adventurer.

This way to work doesn't scale well. People should have other (easy) ways to 
choose a client than asking their neightboors.

> Though client reviews won't hurt. Contribute them to Linux Journal or
> something. It would get the word out too which results in people doing
> what's described in the above paragraph.

Still, probably a bit too technical : we would miss our target.

> > B. do something like the Apache Software Foundation : best
> > clients/libraries/servers should be endorsed by the JSF, so users would
> > be able to choose between a restricted number of
> > clients/libraries/servers. Of course, criterias for the selection should
> > include things like : - the code must be documented.
> > - the licence must be free.
> > - stuff like a public CVS, a bugzilla, etc ... must be available.
> 
> See jabberstudio.org. It could be improved to point people towards specific
> projects. I know SourceForge.net does more than just host projects (ie:
> Most active, popular, etc.).

See http://www.apache.org/. 50% of the projects on jabberstudio.org are
either unmaintained or dead. Don't get me wrong, this is great : people
do experience with Jabber a lot. But still, people don't want to test 20
clients before finding one that suits their needs. So a list of usable
clients/libs/components/servers, with their features, would be great.

> > All those ideas are non-technical work. I think that technical work has
> > proven to be unsuccessful with solving the jabber community problems.
> 
> I'll beg to differ on that point. The lack of technical work is the problem.
> If tech work was unsuccessful, stpeter's mantras wouldn't all be "...more
> code". They would be "Less code, more process". Anti-mantra #1. :-)

To summarize :
Problem: "There a too many unusable or undocumented clients/servers."
Solution: "Let's do another one !"
Does it sound right ? 
-- 
| Lucas Nussbaum
| lucas at lucas-nussbaum.net    lnu at gnu.org    GPG: 1024D/023B3F4F |
| jabber: lucas at linux.ensimag.fr   http://www.lucas-nussbaum.net |
| fingerprint: 075D 010B 80C3 AC68 BD4F B328 DA19 6237 023B 3F4F |