incubator-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marvin Humphrey <mar...@rectangular.com>
Subject A lot of bodies, a lot of money, and a source-dump release.
Date Tue, 07 Jun 2011 01:40:56 GMT
Greetings,

After a long period of reflection, I have accumulated many serious concerns
about the viability of the proposed OOo podling.  There are a lot of unknowns
which make it hard to predict whether the project will become self-sustaining
-- but while the grand experiment plays out, it is going to put a great deal
of strain on the ASF's resources, institutions, and community.

First, it has been pointed out that the ASF does not have significant
experience with end-user applications.  We will have to thrash out how to do
end-user support well, and I expect many painful mistakes along the way which
damage the goodwill associated with the Apache brand.

Second, others have noted that the ASF is accustomed to making source
releases, that figuring out how to release thousands of binaries is going to
be hard, that our current QA resources and traditions are nowhere near
adequate, and that it is doubtful that our network of mirrors is up to the
task of distribution.

Third, everyone acknowledges that this huge, complex, old code base is going
to require a very large community to sustain it on an ongoing basis.  I am
satisfied that the current list of initial committers achieves a minimum of
diversity to suggest that the podling has a shot, but we are way, way, way
away from what it would take to graduate and become a healthy top-level Apache
project.

Unfortunately, given the sordid history of OpenOffice.org, expanding the
community is going to be difficult.  A lot of volunteers who have worked hard
to achieve proficiency with the software are committed to the competing
LibreOffice fork, which has a divergent codebase.  A second crucial talent
pool -- Oracle's employees -- also appears to be off-limits, except in
managing the transition.  We are left with IBM to provide the bulk of the core
dev expertise -- which would ordinarily be fine for an Incubator podling, but
the scope of this project makes it a special case.

Sadly, in my opinion early outreach has been hampered by a sustained series of
impolitic communiques on the part of certain project personnel, which continued
even after repeated guidance from ASF veterans.  Improving understanding of
open source culture is part of incubation and there is always time for
redemption, but the fumbled launch of the recruitment effort which is so vital
to this podling's survival has dealt it a cruel setback.

The proposed podling also has to be prepared for the pullout of IBM at any
moment, Harmony-style.  Business is business, and this is what the ASF signs
up for by being a commercial-friendly organization -- we need to be wary and
hedge our bets.  An OOo podling/project is necessarily going to be
extraordinarily reliant on the expertise of paid developers.  The ASF has
ample experience with the instability of such arrangements, and it should read
negatively on this proposal.

Given all these challenges, I believe that the proposal needs substantial
improvement, and that it will have to be IBM who steps up.

First, we need to see a lot of bodies.  

By necessity, these will come from IBM to start with, and I see that there are
now five individuals on the wiki listing IBM as their affilication.  My
seat-of-the-pants target is that the initial committer ranks should clear what
is necessary to publish and support a first consumer release by a wide margin.

Second, we need to see a lot of money.

I would like to see a budget drafted with the assistance of the Infrastructure
team spec'ing out machines, bandwidth, etc, which should exceed the predicted
requirements by a comfortable multiple (3x-10x), and in a configuration which
caters to the expertise of existing Infrastructure staff.  I would then like
to see a binding committment from IBM to fund this budget -- with cash, not
hardware donations.  It seems to me that proceeding in phases would be fine,
but the ASF must be ahead of the game at all times to account for a potential
podling-killing IBM withdrawal announcement.

It also seems to me that an increased donation from IBM to the ASF general
fund would be appropriate, considering the administrative, PR, legal, and
project-management costs of swallowing this enormous beast and all the
bitterness that has attached to it over time.

Lastly, I would like to see the proposal's backers comment on the possibility
of setting an early goal to deliver an IP-clean ALv2 source-dump release --
basically, the cleaned-up code dump that Greg Stein has pondered (which may
prove trickier to deliver than we anticipate -- are there shortcuts to
rewriting around problematic dependencies?).

I think there's a substantial chance that this podling will not make it through
to graduation, and it may not even make it through to a successful consumer
release depending on IBM's stamina and business interests.  If it doesn't make
it, I don't want the ASF to have absorbed mammoth opportunity costs and
volunteer time without getting anything in return.

Thus, in my mind it would be good to see the podling prioritize the publication
of an IP-clean source dump which can be strip-mined by IBM, TDF, me, you, or
anybody else under the terms of the ALv2.  It would be nice if it built and
ran, but to my mind that's a lower priority than just putting something out
there that people can scavenge with confidence.  

Once that's done, then we shall see if the podling can muster the heroic
endurance that will be needed to launch a successful consumer release, and do
it in a timely manner -- but with the source-dump release in the podling's back
pocket, we can still be pleased with what has been accomplished even if the
end-user effort stalls.

For reasons articulated by Ralph Goers[1] and Bill Rowe[2], I generally favor
giving the OOo proposal a chance, but I also believe that it is accompanied by
exceptional costs and risks to the Foundation which need to be taken into
account.  In theory, I would like to see companies such as Oracle encouraged to
open-source valuable software through us under a permissive license, but I'm
displeased that we're to take ownership of something stained by buckets of bad
blood, and it doesn't make sense to accept the gift at any cost.  I also want
to say "yes!" when companies such as IBM propose to work within our framework,
but I hope that if we say to them that while we greatly value our existing
working relationships with IBM employees on other Apache projects, this one is
so messy, costly, and risky that must decline, they will understand.

Therefore, if those three concerns -- bodies, money, and a source-dump release
-- cannot be addressed, I will regretfully vote -1 to deny the OOo podling
entry into the Incubator.  

Marvin Humphrey

[1] http://markmail.org/message/op7svuogunsmkyvz
[2] http://markmail.org/message/dgzovrdcheeo3dhd


---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Mime
View raw message