Nobody has yet posted any re****t of the Ingres Engineering Summit anywhere,
so I thought I'd re****t a few bits and pieces that captured my attention.
First of all, it was an open event, as befits an open source company.
Anyone was welcome to attend and a great many partners (and others) did.
I
don't know the total headcount but considering it was a moderately
expensive
week and a long time out of the office, it was very well attended. I'd
guess there were well over 100 people there.
The amount of openness demonstrated was impressive. It wasn't just about
exposing the code, it was also about exposing the thinking, the working
practices, the people, and--yes--the tensions. With the exception of a
one-hour slot that was employees-only, everything was out in the open.
Anyone who didn't attend has missed out on a lot. Plan to attend next
year.
Just running through the agenda from top to bottom, and focussing just on
Ingres rather than OpenROAD, here are some highlights:
Andrew Ross did a one hour presentation that covered a lot of ground on
the
general theme of "community development", and more precisely, the barriers
to community development and what has to change to make it
easier/possible.
As we all know, there was a lot standing in our way, and there is a long
way
to go still. However the need for a properly open process was as taken as
given. As well as re****ting a lot of progress with things like
http://code.ingres.com
(Subversion code repository), http://lxr.ingres.com
(code cross-referencer), and http://bugs.ingres.com
(bug tracking), Andrew
also outlined a number of as-yet unresolved problems. I think I'll leave
it
to Andrew to elaborate on those, but the big one (IMO) is the
communication
channels. Mike Sale, Mike Leo and I have been discussing this too. One
thing we agree on is that the current phpBB-based Ingres forums are an
embarrassment and have to go (real soon if I have my way). Mike Leo's
suggestion of using vBulletin turns out to be top of the list at Ingres
too,
so that could happen fairly quickly. There are several benefits to
vBulletin but the big one is that it will allow us to have a single
community delivered via web pages, e-mail, or NNTP, so that we can
accommodate everyone's way of working. IRC will of course remain
separate.
Another big item from Andrew's presentation is that there are actually two
fairly successful virtual development systems about ready for delivery.
We
should be able to get our hands on these within a week or so.
I won't dwell on the rest of the presentations in such detail. You can
infer the significance of these topics appearing on the agenda as well as
I
can.
We had a number of demos of things like Ingres Café and some OpenGIS
software. There was a presentation from Gordon Thorpe on the formidable
challenge of re-architecting GCA. Hugh Darwen spoke about Project D (an
implementation of Tutorial D on top of Ingres) and earned himself the
second
prize for Best Presentation. We had two presentations on column stores,
one
of which went on to win the first prize Best Presentation (Marcin Zukowski
on Monet/x100). Karl spoke fluently about something or other for an hour
and many of us marvelled. Mike Touloumtzis led a discussion about how to
implement column encryption; nothing was decided but a lot of ground was
explored. Steve Ball and Alison Stillway led another discussion on how to
implement MVCC and which model of MVCC to adopt; nothing was decided
except
that a design do***ent will be drafted for public comment. (To my mind
this
may be the most immediately and widely useful enhancement that popped up
on
the new-features radar.) Emma McGrattan picked up where Andrew Ross left
off. The big thing in her presentation was the carefully expressed and
several-times repeated instruction that Ingres Corp requires all its
personnel to devote 10% of their time to community projects (i.e. 1/2 day
per week). That is a lot of effort folk! After Emma, Kai-Uwe Sattler
talked about some research into making Ingres more autonomous and
self-tuning (including the ability to recommend secondary indices and
statistics). There were two presentations on two different replicators
from
partner companies. Roger Whitcomb told us about the work he's been doing
on Ingres Management Tools, which was really good stuff. (I was present
at
the meeting where VDBA was first unveiled back circa 1995 and it was
greeted
with horror and revulsion then, and nothing has changed. Roger's work is
definitely going in the right direction this time.) After this I stepped
out of server-land and saw Daryl Monge discussing Ruby on Rails. For some
reason every head in the room turned to look at me when he reminded us
that
RoR requires every table to have a synthetic integer key. Evidently
everyone understands this is wicked and wrong and that they should feel
guilty about it, but equally evidently people are just going to go on
doing
it anyway. That was enough for me, so I retreated back to server-land
again
after that.
So there you are. I saw fewer than half the presentations, so perhaps
someone else will comment on the others.
Finally , I am hoping to get at least a couple of these repeated at the
IUA
conference on June 17 in London. Please let me know if there is anything
above that particularly takes your fancy and I will see what I can do.
Roy
|
