Application certification (was Re: [Beowulf] hpl size problems)
Robert G. Brown
rgb at phy.duke.edu
Sat Oct 1 12:12:15 PDT 2005
Andrew Fant writes:
>> I think you're saying that RH is the new MSFT, which seems pretty funny...
> Actually, I see it as rather sad. As it was pointed out in a parallel thread,
> Gaussian will refuse to support your installation if you are using anything
> more recent than RedHat 9 and even them, you aren't supposed to put security
> patches on. To too many ISVs Linux := some version of RedHat. I will admit
> that I am somewhat of a Gentoo partisian, and that makes me suspect. But why
> should I be expected to mirror my dependency trees into an alien db format and
> jump through hoops to compile a backporting nightmare of a kernel just to
> support "standard" code? Red Hat does not own Linux, even though 9 out of 10
> CIOs would be hard pressed to name another distribution. The whole point of
> Open-Source software is supposedly about being able to get your hands dirty
> with the code and have meaningful standards for interoperability that one can
> reproduce without being at the mercy of ANY commercial vendor.
You're not "required to". You only are required to if you want your
apps to be able to build and transparently install and run on
LSB-compliant systems. However, linux and open source is all about
freedom, right? You don't have to do anything.
Of course if you DON'T try to meet some sort of ABI, API, packaging
specifications, chances are excellent that your application will not be
terribly portable, and that it just won't run on many/most linuces that
are out there without a lot of work. That work, in turn, is expensive
in every sense of the word.
The thing about standards or ATTEMPTS at standards like the FHS and LSB
and Posix and ANSI and the ODF are that they ultimately enable a lot
more than they restrict, and even the "Walrus of the Day" that dominates
the standard development process as standards are introduced has to play
on the same field as everybody else once they ARE open standards. This
is where MS makes up the rules as they go along, and why there ARE no
meaningful standards in MS development -- there are only current
practices that will magically change before your very eyes the first
time MS notes that a competitor is out-engineering them on top of their
existing "standards". Even REAL standards aren't safe from being
co-opted and shifted -- with the market clout that they have, they can
muscle people who actually comply with standards out of the market
altogether by not supporting them properly (so the competitors' code
breaks on their systems) and adding functionality that is proprietary or
hooks into proprietary that is outside of the standard. MS regularly
does both and it has made them the largest and (in many senses)
wickedest monopoly the world has ever seen.
Can RH do this? Of course not. Really, don't make me laugh. They
Don't Own The Code. Not even the package code. Everything is open
source, open standard, and subject to a ruthless genetic optimization
process -- if it isn't sufficiently portable, isn't sufficiently
powerful, it ain't agonna fly. Look at up2date. Does anybody use it?
Only if they are a brain-dead company forced into it by their license.
Everybody else uses yum. Red Hat uses yum. Mark was right -- here is
an example of a tool that works on TOP of the rpm libraries, is (I
assure you) independently developed from Red Hat and although it is now
being co-developed with rpms and debs and other things that is BECAUSE
the many linux distros realize that it is in their own best interest to
eventually make the LSB happen. If they fight among themselves over who
gets to be "The Linux", the commercial software developers of the world
will never view Linux as anything more than a nuisance and support
This matters. Linux is arguably right up against MS at the desktop --
Open Office vs MS Office, an abundance of powerful and attractive mail
clients, a grab-bag full of powerful applications of all sorts for doing
this and that. For programmers and serious computer types it is the
system of choice. For normal humans, though, it still lags. Why?
Not because of media hype, not because of MS FUD, not even because MS
has a stranglehold on the system vendors. It is because:
a) Vendors can sell hardware with enclosed (possibly non-open-source)
drivers that will actually work Sure, this is something of a chicken and
egg thing -- they support MS because MS is ubiquitous in their market
which keeps MS ubiquitous in the market -- but Linux is designed to be
deliberately UNfriendly to third party and closed source software. The
kernel taunts you with warnings of taint if you insert such a driver.
b) Even on friendly turf, the rule for most linux packages or
applications is "rebuild it on the distro you'll run it on" because
otherwise if I try running an RPM from distro X on distro Y, chances are
far too great that I'll either screw something up with e.g. dependency
loops, wierd obsoletes, unexpected paths or it just won't work right or
both. Linux eventually has to get to the point where the whole thing
doesn't have to be rebuilt, from source, after patching, in order to
obtain a consistently installed and functional binary package. WinXX,
OTOH, still runs DOS applications from twenty years ago. Not strictly
fair, of course -- WinXX has its own problems with shared libraries and
dependencies -- but one can still often get a commercial application
written for e.g. Win95 to run on WinXP.
c) This makes it possible for software companies to write for WinXX
with the possibility of their work having a lifetime without major
revision and rebuilding required of over a year. They only have to
worry about getting it to run on ONE standard, as well, not a half dozen
(each of which has too little market share to be worth it on their own,
even if Linux in aggregate does). They can't just sell "Linux"
software, they have to sell "Red Hat Linux" software, or "SuSE"
software, and force the user to make the choice to use what they
require rather than face the Evil of Heterogeneity (for they too follow
the rules I laid out yesterday and consider Heterogeneity an Expensive
Evil rather than a Good Thing.
So <sigh> yes, you don't have to play, gentoo doesn't have to play,
Debian doesn't have to play, SuSE and Red Hat and Mandriva -- nobody has
to play. Each can continue to do things "their way" (each for the best
of reasons, of course) and the state of affairs above will persist.
At most 2/3 of all hardware vendors will support linux, so e.g. a huge
number of printers will be too difficult for most users to get to work,
gamers won't be able to use 3d FX on their high end video cards so game
playing will continue to suck so much that Blizzard etc won't bother
releasing native linux versions of their games etc. Software companies
won't write software for linux because they won't be able to sell it
much if they do and because -- which linux will they right for? They
can't support it on all of the, because they are all very different.
Desktop users will continue to choose Windows because there they can see
an application or game or camera or printer, buy the application or game
or camera or printer, take the application or game or camera or printer
home and install it (usually) themselves, and (usually) end up with a
functional application, game, camera, or printer. If they try that with
linux, they'll need to have a linux genius handy to get the application
(written for e.g. RH) to run under Debian, to get the game to run under
Cedega or Winex (if they have one of the cards that can be binary
supported under linux at the risk of "tainting" their kernel --
otherwise performance will be so bad they can't play it anyway), to
search the web to find an obscure site where some geek reverse
engineered the camera's image encoding and handshaking so they can badly
retreive the images and manipulate them, or to find a printer driver for
an earlier version of printers made by the same company and hope that
interior stuff hasn't diverged so much that it won't still work.
I personally devoutly hope and pray for the LSB to slowly, ponderously,
evolve into a real standard and for a certification consortium to emerge
out of this that blesses the essential linux core, certifies software
commercial and distribution and otherwise against that core plus
certified midlayer components, with at least the major linux distros all
on board and cross-distro compatible in maybe three years (which would
be quite fast in this business. That would make everybody's job easier.
All the distros will obviously have to give up a little bit of their
"identity" to play, but the long term benefit will be that maybe, just
MAYBE, linux will start to attract box-set software sellers' attention,
games software sellers attention, become a platform one CAN write
drivers for (open or closed source) without open hostility if they
aren't open and contributed to the kernel source (already one of the
most bloated things on the planet).
Seriously, this stands very much in the way of World Domination. I
detest WinXX and MS in general, I think Linux is so much better in so
many ways, but it is not EASIER, especially for really ignorant home
users. FHS, LSB, API, ABI -- these acronyms form a powerful charm.
There is some mighty stuff standing in the wings or already on the table
on the linux side, stuff WinXX cannot afford to replicate. Diskless
everything being one of the primary ones. Repo-based installs and
maintenance. Key-controlled, internet based software distribution --
boxless software marketing. Much of this is NOT all Red Hat -- quite
the contrary. But it won't happen if we don't all share this particular
vision and work to make it so, overcoming BOTH the technical problems
AND the economic/corporate problems as we do so.
> Andrew Fant | And when the night is cloudy | This space to let
> Molecular Geek | There is still a light |----------------------
> fant at pobox.com | That shines on me | Disclaimer: I don't
> Boston, MA | Shine until tomorrow, Let it be | even speak for myself
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: not available
Url : http://www.scyld.com/pipermail/beowulf/attachments/20051001/1de99d50/attachment.bin
More information about the Beowulf