Subject: Re: more vs. better (was Re: morning hack...)
From: Paul Rohr (paul@abisource.com)
Date: Tue Apr 17 2001 - 16:23:25 CDT
At 12:19 AM 4/13/01 -0400, Dom Lachowicz wrote:
>Feature freeze has always been on the horizon... well, at least for the
>features listed on the AbiSource page that haven't been touched in a year or
>two.
That page, unfortunately, was an occasional burst of inspiration on Eric's
part that otherwise got ignored. Replacing it with something accurate and
current would be wonderful.
>What I'm saying is that this feature matrix needs to be re-evaluated, combed
>over, pruned, extended, and then implemented.
I agree wholeheartedly with all but the "extended" part.
It's one thing if new developers show up with sneak-ins, because we can
always release without their work, since nobody was expecting it. The bar
they have to clear is very very high, though. Any such sneakin which might
destabilize the tree and doesn't clearly Just Work can just sit out there as
a patch until we open the post-1.0 tree.
The idea which makes me itch is seeing core developers starting new features
instead of finishing what we've already got.
>But our list goes further. Martin posted a TODO list that we came up with at
>GUADEC. I want this list to be fluid, at least for the next week or two. We
>can all evaluate what needs to be done, and what we'd like to see in the
>release.
Yeah, I still need to read that list. After I finish wading through raw
email, I'll start opening attachments. (Sigh.)
>I'd dislike to release a feature-disadvantaged 1.0 just about as
>much as I'd dislike releasing a bug-ridden (or 1/2 implemented features)
>1.0.
>
>If you see my point, I commend you, because it's a fine line...
Feature-disadvantaged compared to what? Every release we do is better than
the one before, but to date there is *no* stable supported binary release of
AbiWord out there for anyone to use.
If we wait to release AbiWord 1.0 until it can do a rough approximation of
everything Word can do, then we'll exhaust the patience of a world full of
people waiting for us. I know that this isn't what you (or anyone else) is
seriously advocating, but sometimes, your "fine line" is pretty hard to see.
>>Ask yourself which target you think those millions want us to shoot for:
>>
>> - N features that all work flawlessly, or
>> - N + 5 features that all work pretty well most of the time.
>
>I'd like N + 5 features that work flawlessly. My mother always said that
>some good things are worth waiting for. This is the only time that I've ever
>agreed with my mother ;-)
OK. Time to put a stake in the ground.
I'd like to see us ship a *rock-solid* implementation of the features we had
*prior* to GUADEC, and call it 1.0. Am I the only one worried that we'd
have trouble doing so by the 3-year anniversary of AbiWord 0.1.0 -- namely,
this August?
Wouldn't you say that our original AbiWord users have been waiting long
enough for *this* "good thing"?
I need to be convinced how we can meet or beat that goal before I'm going to
have much patience with the idea of delaying 1.0 for even more stuff, no
matter how cool.
>>Opening up the flood gates to even more features just postpones this
>>much-needed work, and delays our eventual release indefinitely.
>
>Not entirely or necessarily, but your point is well taken. Human nature and
>all...
OK. Sounds like we may be able to arrive at a mutually-satisfying solution.
>>Are you proposing a Linux-style even numbered stable release that we'd
>>devote resources to maintaining as such? Is there a group of people
>>willing to create, support, and maintain such a release?
>
>Not entirely, I'm advocating not skipping the 0.8.0 series. Anything before
>1.0 should be considered development and unstable. Point releases are
>usually considered stable/unstable after a 1.0 release. n < 1.0 means
>alpha/beta code, so beware, even if the middle number is even...
To date, all Abi releases have used odd middle numbers, though. The change
in precedent wouldn't be obvious to many folks.
Since we're all agreed that 0.9.x should correspond to a rough feature
freeze, I'm starting to suspect that the reason you may want a 0.8.x series
is so that we can fold in another round of new features before 1.0. Am I
understanding you?
If so, my suggestion would be that any and all the features you're tempted
to develop during the 0.8.x cycle instead be developed during a 1.1.x cycle
instead. The sooner we invoke the feature freeze and get through the 0.9.x
stabilization prior to 1.0, the sooner we can get rolling on 1.1.x fun
stuff.
>>Paul,
>>designated grinch
>
>Not a grinch, possibly more of a "voice of reason", "second opinion", or
>"devil's advocate". You gotta learn to choose your words more wisely :)
:-) VOR sounds too arrogant, SO is definitely too modest, and DA is more
confrontational than I intend to be. My choice of DG may not be "wise", but
it was deliberate.
For some people, Christmas is the day you get to play with lots of cool new
stuff. If you're a resident of Whoville, that's a really fun day, so long
as someone delivers that cool stuff. However, I'm the one who's insisting
that if you want your products on store shelves for the Christmas season,
you need to freeze it for your OEM channels no later than mid-summer.
(Literally.)
On second thought, I guess that makes me the *least* grinchly person here,
huh? I want to ship the AbiWord 1.0 we can have for Christmas *this* year.
Please don't make us all wait for Christmas 2002.
Paul
motto -- finish and ship, don't start or wait
This archive was generated by hypermail 2b25 : Tue Apr 17 2001 - 16:16:04 CDT