[Mod-pubsub-developer] Re: Do commercial deployments count?

[Kragen S. <kra...@ai...>]

I'm sending this as email rather than posting this as a blog comment
because the 40x10 text box on the blog is a pretty terrible user interface
for editing a 700-word essay, because it doesn't facilitate citation
the way email does, and because this way, I get to keep a copy
automatically.

Joyce Park writes at
http://www.mod-pubsub.org/blog/archives/377_Do_commercial_deployments_count.html:
> I've been turning over in the my mind the question of whether commercial
> deployments are the proper measure of an Open Source project's success. We
> get asked more and more frequently by businesses for benchmarks equal
> to those put out by commercial operations. Also, we get asked whether we
> have commercial deployments already (we do, if it matters); and potential
> users mention whether they will or will not be referenceable accounts,
> like that should be important to us.
> 
> Maybe this is inevitable, given Mod-pubsub's genesis inside a
> corporation. But when we started this project, we specifically did not
> want to compete in the enterprise market. We wanted to serve the Web
> developers of the world, especially in the social software space, in the
> same way that products we admired (Apache, PHP, Movable Type, wikis) had
> started out. This accounts for (to cite an obvious example) our decision
> to tie our product to Apache httpd despite its performance-limiting
> aspects. 

What is this about being tied to Apache?  The Python event router isn't
tied to Apache, and never has been.  I wrote it in the first place largely
because our salespeople didn't have Apache on their Windows laptops. (And
because I was frustrated with the lack of progress in the actual product,
a lack of progress that apparently continues at KnowNow to this day.)
In the product we're deploying pubsub.py in, we have Apache, but as it
happens, Apache and pubsub.py never talk to each other.  (Yet.)

So in what sense is mod_pubsub "tied to Apache"?

I agree that a full-featured HTTP server is useful in that it lets you
do proxying, SSL, and authentication when you access the event router,
and hack around browsers' "security" models, but it's not clear to me
how Perl or PHP integration or URL rewriting benefits the event router
itself (as opposed to ESPs that access it.)  And what's forwarding?
You mean generating HTTP redirects?

> We thought this decision would give our project some much-needed
> focus, make a virtue of necessity in our lack of enterprisey features,
> and bring the joys of pubsub to the largest pool of users. Also, we
> deliberately chose the BSD license knowing that it meant users are not
> obligated to return their changes to the community.

I really wish mod_pubsub were under the GPL, because right now, it's
pretty tough for me to get open source enhancements out of AirWave,
even though maintaining the enhancements on a fork costs us more in the
medium and long run.  If it were under the GPL, we'd have to license our
enhancements to the public under the GPL anyway, since we're distributing
the software, and sending you a copy wouldn't be a big deal.

So if you want enterprisey features, the GPL is probably a good way
to get them, unless you expect a C2.net to pop up and start selling
pubsub.py+replication or some similar product as proprietary software.

Too fucking bad KnowNow released the code under a GPL-incompatible license
in the first place.  You could still license further improvements under
the GPL with a special exemption for the advertising clause.

> So now we're at a bit of a crossroads: should our big push for 2004 be
> to make it easier for newbies to dig into Mod-pubsub -- at the cost of
> basically telling the enterprise evaluators that we don't have time for
> them? Or should we try to focus on the things we need to do to make the
> system more scalable and performative? Give us your advice please.

I think we should build as many applications as possible with mod_pubsub,
and encourage other people to do so, too.  I don't know whether blogging
hobbyists or ISVs are likely to help more with that, but I suspect it's
the hobbyists.

Mark Baker writes:
> newbie focus, e.g. packaging. and if you think that won't help
> enterprise pickup, well, yer wrong. 8-)

Yes, of course, packaging is important for adoption at folks like
AirWave, too.