The last talk yesterday at ApacheCon, was Roy T. Fielding presenting
HIS (no consensus yet) "call for arms" to move on to a Apache Server
3.0. For those of you who doesn't know Roy, he was the author of the
HTTP specs and a co-founder of Apache. He also wrote a thesis paper on
the Representational State Transfer architectures, more well known as
RESTful architectures.

There were many interesting bits (you can probably find the
presentation online somewhere, but what strike me the most were;

 * Apache Httpd 2.x development has slowed down to a crawl.
He had analyzed mail traffic since the start in 1995 and correlated it
with releases. His conclusions are roughly;
 - People only volunteer work when there are a lot of things to do.
 - One Leader will "Lead by Doing", and if people agree they follow,
otherwise the leader is no Leader.

* Apache Server project needs to focus on supporting the users  - - - -   NOT.
Apache is not a commercial company, and should not focus on giving the
users what they ask for, but to develop what the developers need, find
useful and think is FUN enough to actually do it. Apache 2.x is right
now not fun.
This means that Apache Server 3.0 (the "httpd" part will be dropped
from the project), will have "kiss my ass" compatibility with 2.x.

* Roy has introduced a token-based protocol and API called "waka",
which will be used internally. All incoming requests will be
translated to "waka://" at entry point of a system, and translated
back at the exit point to whatever the protocol is. It will also mean
that the much more efficient waka:// will be available from Apache
servers, although not an industry standard, it might well become one
thanks to Apache's dominant role.

* There will only be a single Process Model per supported platform.
The many MPMs, pre-fork, threaded, event and what not, will be reduced
to one per platform.

* No compile-time switches. Every shipped Apache Server will be
identical on the same platform.

* No default configuration, instead use useful hard-coded defaults and
command-line switches. Developers should get something useful by
simply starting "httpd run" from commandline, without any config
files. Also, configuration file format to be less cryptic.

* Internal changes making it a lot easier to create modules.
Currently, every module seems to need to understand the entire
internal architecture, and can not focus on small sections. He gave a
good slide on HTTP headers, which actually are something like 6
different types of data. Find that to ponder over it.


* Finally, and this is the MOST interesting bit; Roy is also trying to
open up the Apache Httpd project's sandbox to all Apache committers,
and make a very low bar of entry for external people. He thinks it
will stimulate innovation and attraction of the project, needed to get
this Apache Server 3.0 off the ground. This is HUGE. The httpd project
is extremely conservative, 6 months of continual submission of patches
before you come up for committer status discussions, and
"Review-Then-Commit" policy on all changes to code. So, if YOU are
interested in this Apache Server 3.0 project, join the discussion at
the [EMAIL PROTECTED] mailing list, voice your support for an "Open
Participation" community model.


Cheers
Niclas

_______________________________________________
general mailing list
[email protected]
http://lists.ops4j.org/mailman/listinfo/general

Reply via email to