[nSLUG] Jigdo File

Ben Armstrong synrg at sanctuary.nslug.ns.ca
Wed Mar 5 18:00:08 AST 2003

On Wed, Mar 05, 2003 at 04:12:30PM -0400, Donald Teed wrote:
> Commercial software development has been around for a long time and
> has established traditions for what version numbers mean.

I am sure those traditions make sense within that development model in 
order to successfully meet the needs of their market.

> The way Linux has developed a system with odd and even point kernel versions
> is a cool enhancement on these traditions.  Putting out software
> with versions *far* less than 1.0 and saying it is "useful" or maybe
> even robust, but please check the mailing lists, bugzilla, and google
> and everyone you know first, is not a practical expectation.

So what of my alibi project?  It *is* useful and robust enough for a subset
of users.  If you are a potential alibi user, but have needs that fall
outside of those of my present small user base you may or may not be served
well by my project.  However, I do endeavour to make each release a
complete, finished release, usable for at least a subset of users, and as I
progress from milestone to milestone I expect to be able to grow that user
base, gradually meeting the needs of more people.  All I'm using the
version# for here is to say "this may not serve the needs of all of my
itended users for my one-point-oh release".  I'm not using it to convey
stability or bug-freeness.  It is strictly my idea of the "finishedness" of
the first version of the project.

I don't have the resources to hide away and not release alibi until it
supports everything I want it to support for my one-point-oh.  At some point
I have to release.  My first public release is at 0.1 because I need the
experiences of users outside of the small base of testers who are helping me
get the early versions off the ground.  So I release it with caveats: "this
only supports blah and foo so far, and I don't guarantee it will work for
you if your needs are different".  And that is typical of newly-created free
software.  Even if not explicitly stated, I believe this sort of disclaimer
is implicit.  That is *not* typical of commercial software.  They have to
hold off and not sell pre-releases (although they may give some away).  They
need to clearly define who their market is and they *must* release a 1.0
that meets the needs of enough people in that market to keep their company
afloat.  They must also release it in a reasonable timeframe, or they will
go under.  With free software we have the luxury of releasing earlier and 
delaying one-point-oh longer, so we can "release when ready" instead of 
being subject to the whims of a commercial market.  It's an entirely 
different world.

Or to put it more bluntly: "alibi 0.1 would be a flop as a commercial
product; but as a free software project, it is already a success".  I 
believe what is true of alibi is already true of jigdo too.

> When the only means to obtain a CDROM within a few hours of time (as
> I grew to expect with Redhat) is a mandatory server technology that
> is unreliable, there is a problem with a too early adoption.

I don't think so.  I'm hearing from i386 jigdo users that it works very well
for them.  Too early adoption?  Apparently the subset of potential users who
use already use jigdo are happy enough with it.  Your platform (sparc) was
officially supported in Debian before jigdo was available, afaik.  What did
sparc users do before jigdo?  Or what do they do today?  Did you ask the
debian-sparc list about what they use?

The Debian ports web page tells me sparc existed since Debian 2.1, three 
stable releases ago (an eternity! ;)  The debian-cd mailing list tells me 
that jigdo was "not ready for release yet" as recently as May of 2002.  So 
there's a disparity here between "jigdo is the only way" and the recorded 
history of the Debian sparc port.

> The Debian website said it was the best way to obtain ISO images.

For most people it probably is.  I am sure I don't need to remind you that
most users use i386.  Furthermore, among those who don't use i386, I haven't
yet to find evidence that other non-i386 platforms have problems with jigdo. 
You quite possibly are one of the first users to run up against your
particular problem (jigdo + woody + sparc == trouble) or maybe your problem
isn't even that (jigdo + bad mirrors == trouble) as hinted at by your
closing remarks.

Generalizing from your particular bad experience to "jigdo is unreliable
software that was released to the public too early" is quite a stretch.

> For installing from CDROM images, there are two options: fetch
> full CD images, and assemble images using jigdo.  I tried many
> mirrors and no one carried ISO images for sparc.  I believe
> the Debian website warned that finding full ISO images would
> become difficult as more mirrors adopted jigdo.

Again, I wonder what the debian-sparc community has to say about what the 
best way to install Debian for sparc is.  I do know from looking into alpha 
and ppc support that if you go off the mainstream (i.e. non-i386) expect 
glitches, because many developers of all the pieces of software that go into 
Debian have only i386 systems themselves and therefore non-i386 platforms 
are tested less rigorously.  This means that part of the burden of testing
and debugging trickles down to the users themselves.  Unfortuante, but 
hard to avoid, given the structure of the project.  We don't have the 
resources to employ a lab full of non-i386 testers who thoroughly test every 
aspect of Debian before it is released to non-i386 users.

> If I want to install today, I can't wait for CDs in the mail.

If I were in your position, I'd be inclined to ask the Debian sparc user
community simply: "Jigdo doesn't work for me.  How do I obtain sparc CD
images to download?  Or how do I make Jigdo work for sparc?"

> My sparc box had no floppy drive.  I didn't see bootable
> CDROM boot disk images for sparc as there were for i386.

I'll admit, that puts you in a bit of a spot.  When I first posted in this 
thread I wasn't aware of those constraints.

> In general I like to obtain CDs because they always work,
> regardless of the network access in a particular location.

Sure.  I understand the desire to go with CDs, especially in your situation.

> The bootp option was explored and we didn't have the ability to
> add custom lines to the DHCP server (config generated by
> a script they didn't want to change just for this).  Eventually we
> found that if rarpd was running on the same machine as providing the
> tftp, this worked for the boot image.  I posted that tip to
> this list in case it might help anyone else out.  (We do
> get googled)
> If that option had not been available, I would be stuck with jigdo.
> I would have likely dropped Debian and moved on to something else.

I still don't see this as a complaint about the fitness of jigdo for general
consumption, but rather as a specific bad experience with a fringe
circumstance: apparently there are Debian sparc users out there, and they
installed their systems somehow.  But your expectation was that jigdo would
work on sparc as well as it apparently does on i386 and you were let down. 
If you poll debian-sparc and find people saying "yup, jigdo works for me"
then you know you've hit a bug or environmental problem (like a broken
mirror) that doesn't affect users generally.  If you poll them and find that
they say "no, don't use jigdo" then you use whatever they recommend instead.

> > > "We" is the entire community of open source software users.  If the
> > > users don't exist, the developers might as well be playing Starcraft.
> > 
> > So you are now a spokesperson for the entire community of open source
> > software users?  I'm sure you have some good ideas about what users need,
> Why not?  You are a spokesperson for Debian developers are you not?
> It doesn't really matter.  We can pretend, just like the people
> on TV pretend that popular opinions matter.

OK, let's pretend that your opinion represents that of the mainstream user. 
The users of Debian clearly do exist.  And I'm pretty sure the developers
aren't just friterring away their time building stuff that doesn't meet
their needs.  Now, where is the great hue and cry from the user community
about what a bang-up job the developers did on jigdo and that it was
released prematurely?  Hint: try the debian-cd mailing list.  I did a quick
survey of the the past four months of the list and couldn't find anything
substantial in the way of complaints.  Mostly what I found was praise and
thanks.  I did find one thread in which one person was displeased, but the
summary of his complaint is "jigdo isn't as good as wget" to which the quite
reasonable reply was: "well then, use wget". :)

> I'm not about to cross examine the only apparent link I have
> to getting a Linux installed!  If anything, my next step would
> likely have been to abandon Debian and find another distro for sparc.

OK, what other distros exist for sparc?  As much as I'm a fan of Debian, if 
someone's having a really hard time with it, I'm willing to say "maybe 
Debian isn't for you and you should try something else."  I have nothing 
personally at stake here.

But really, I think that between debian-cd and debian-sparc you should've 
been able to find help.

> When I am trying to get something up using a bunch of methods and coming
> up against delays and walls, I don't tend to take the time to look up
> bug reports and file new bug reports.  If I were home, perhaps I'd go
> that route, but not when I feel that I should account for my time.

That's fine.  I said do that at your discretion.  If you don't have the time 
for formal bug reports, then don't.  But at least check with the user 
support lists to see if your expectations are in line with reality, 
*especially* if you're doing something that is not mainstream use for the 
software you're trying to use.

> If Debian developers are concerned with adoption in business
> contexts, these points should matter to people like you.

Indeed, these points matter to me in a general sort of way.  But on the
flip-side, I'm not paid for my work and I only have limited time to deal
with problems outside my specific area of expertise.  If someone asks me a
question and I find I'm out of my depth, I invariably refer them to the
support mailing lists.  That's one area where I'm proud to say Debian
excels.  You may find me hard-pressed to say whether I care or not if jigdo
+ sparc works for Debian.  I guess I vaguely wish that it would.  But
nothing is personally at stake if it doesn't.  But I *do* care intensely
about whether Debian Jr. works for the people I consider my user audience. 
I'll do whatever is in my power to help them.  Similarly, other developers
within Debian have their own special interests.  So go find those within
Debian who care about your particular problem and see if they can help.

> For my hardware (no floppy) and platform (sun) and network situation
> (no control over dhcp server), it was looking like Debian wasn't
> possible without CDROMs.

OK, so as I said you're describing a very, very specific use of jigdo.

> Installation of the OS is something every system has to go through.
> If that fails, everything else that people put time and energy into
> is moot.  I can understand the problems with bandwidth costs, but
> it seems strange that (a) in the end the same number of bytes are
> going to be needed no matter how you slice and dice them, and (b) other
> places just use mirrors, not special servers to deal with the problem
> of distributing the download traffic.

- Debian is not a corporation
- Debian operates on donated servers
- Debian is operated by volunteers
- Debian's entire offering is huge, far more than can fit on a single ISO
- Debian supports quite a lot more architectures than other distros
- As a consequence, not everyone mirrors every architecture
- Jigdo offers the flexibility to slice out just what the user needs and 
offer it up as tidy little ISO

So certainly we have a different set of needs, so Debian takes a different 
approach at solving the problem of how to get ISOs that contain what the 
user needs to the user.  It's a more flexible, easier-to-scale solution than 
just mirroring a huge complete set of ISOs for all architectures everywhere, 
or trying to guess at a subset that will serve all users for all 
architectures and just offer ISOs that contain that subset.  And most 
importantly, it works for a large enough user base to be considered "ready 
for prime time".

> I am running jigdo in the background at this moment and it seems
> to be better behaved with another mirror.  Sunsite at Ualberta
> must have a botched server or something.  I thought I
> tried other servers when I experienced the failure before but
> it has been awhile now.

Well, here's hoping, then. :)

 ,-.  nSLUG    http://www.nslug.ns.ca   synrg at sanctuary.nslug.ns.ca
 \`'  Debian   http://www.debian.org    synrg at debian.org
  `          [ gpg 395C F3A4 35D3 D247 1387 2D9E 5A94 F3CA 0B27 13C8 ]
             [ pgp 7F DA 09 4B BA 2C 0D E0 1B B1 31 ED C6 A9 39 4F ]

More information about the nSLUG mailing list