Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 06 Jan 1999 23:26:35 -0800
From:      "Jordan K. Hubbard" <jkh@zippy.cdrom.com>
To:        sparc@FreeBSD.ORG
Subject:   All this muttering about support from Sun.
Message-ID:  <613.915693995@zippy.cdrom.com>

next in thread | raw e-mail | index | archive | help
Hey all,

I don't subscribe to this mailing list (I get way too much already)
but I did catch some of the discussion on one of the mail2news sites
and can only say that I find some of what's been discussed so far to
be just a little alarming.  I'm also not sure that my role in
"rejecting" SME's original offer was all that properly characterized,
and the two points are related, so let me see if I can't explain some
of the history here and attempt to put things in perspective:

First off, I'd like it to be well understood by everyone that when you
approach any vendor for FreeBSD-related resources, no matter how big
or small, you're essentially now representing the project as far as
that vendor is concerned and they're just as likely to assume you're
project leadership or something given that most companies are
typically pretty confused about how the whole free software world is
led and who the key players are (I can't say I'm always all that sure
myself).  The danger of misrepresenting or overselling the project is
a large reason why the original deal with SME was turned down.

Even though I myself could probably be said to "lead" the project as
much as any individual does, I am still not at all comfortable with
the idea of representing the project as a whole when it comes to
committing to projects as large as doing and supporting an UltraSPARC
port for any reasonable length of time (a port which flares and dies
quickly being of no use to either party).  Were I to have stood up a
year ago and said "OK Sun, I'll take 3 uSPARCs, $20K in direct funding
and 100 hours of one of your own developer's time in exchange for a
working FreeBSD/uSPARC port", I would have been committing the entire
project to far more work and obligation than $20K and 3 machines will
frankly buy.  It's one thing to do a slap-dash port of FreeBSD to a
new architecture just for our own interest and yet another thing
entirely to produce a port which works to the satisfaction of a
company like Sun, already fussy customers indeed when you realize
you're competing to some extent against the Solaris group on their
home turf.

It was much more likely, given my estimation of our abilities back
then time, that we would have simply failed to produce much of
anything reasonable in the time allotted (oh, did I also forget to
mention that SME wanted the port done in 90 days or less, no-foolin'
at all about the schedule?) and the result would have been Sun
concluding (quite rightly) that we were a bunch of losers who didn't
really know what we were doing.  You just can't say you'll do
something, impose on some vendor's resources (be it in the form of
equipment loans, discounts, cash bribes, *whatever*), and then shrug
with a stupid grin on your face when it comes time for you to come up
with your end of the deal.

We've also shown ourselves to be historically very bad when it comes
to doing things to reasonably timely schedules, the DEC Alpha port
taking 1.5 years from the moment that DEC gave us the machines to now,
where it's actually sort of working.  Such a thing would have been
completely unacceptable to SME and we just got lucky with Digital in
that nobody there really asked us to do the port in any particular
timeline, I also getting the machines through a personal contact at
DEC which made the whole thing much more low-key and low pressure.


So anyway, that's the history up to now and while I'm certainly very
glad to see renewed interest in doing a SPARC port again, I hope that
everyone involved can keep a similar grip on their perspective when it
comes to the more blue-sky aspects of project planning and
committments, either direct or implied, with any 3rd party who might
be attracted to the idea of having a SPARC port of FreeBSD.  It's one
thing to say "rah rah, let's do it!" but a very different thing to
contemplate FreeBSD/sparc actually being committed to -current and
built/released in the same way that FreeBSD/axp and FreeBSD/x86 are;
between where you are now and there lie many gates to walk through,
filters to satisfy, a core team to convince, you name it!

It's also my suggestion that you all start *fist* with the resources
already available rather than hoping that Sun will give you a handout
in the form of discounts or outright loans/grants, something which
will only bring a degree of obligation (nothing comes for free!) into
your little project long before any significant amount of code has
been written or anything is checked into the tree.  That's just not
the way it should be done if you're willing to be halfway honest with
yourself, and the way I've seen every successful porting effort go to
date (observing the NetBSD and Linux folks, among others) has been a
small group of 2-3 developers who have the machines and know-how
(e.g. NOT novice or beginning people) significantly advancing the port
to the point where the other folks can actually jump aboard and start
helping with the more mundane stuff.  Before that can happen, somebody
has to have wrestled with the toolchain issues and a bootstrap
environment (NetBSD/Linux/Solaris/whatever), then they need to
understand what the architecture specific portions of a FreeBSD system
are and get things to the point where an actual FreeBSD kernel is
booting single user on an actual SPARC machine of some sort.

Once you've reached that milestone, some of the people who don't have
pre-existing porting experience can start making more realistic
contributions to the port and advance it from the point where it's a
technology demonstrator to being an actual "product" in the same way
that FreeBSD/x86 is.  Not even FreeBSD/alpha is to that point yet, and
it has a couple of major coding studs like John Birrell and Doug
Rabson working on it.  Just something to keep in mind as you
contemplate this particular mountain, folks. :-)

- Jordan

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-sparc" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?613.915693995>