Re: POSIX shell; bash ash pdksh & /bin/sh
Hi,
>>"Santiago" == Santiago Vila <sanvila@unex.es> writes:
Santiago> -----BEGIN PGP SIGNED MESSAGE-----
Santiago> On 3 Aug 1998, Manoj Srivastava wrote:
Santiago> Thanks. I'm a little suprised that you (more or less)
Santiago> recognize there is a circular argument here :-)
>>
>> Why? I may be stubborn, but very rarely have I been accused of
>> being downright stupid ;-)
Santiago> Mmm, I thought we would agree that circular arguments are a
Santiago> stupid thing.
You have not thought it through, or you are being unusually
dense. You are ignoring time, and the context in which the rules were
created. Bash was made Essential. The reasons do not matter anymore
(we needed a standard shell, there were few alternatives then, bash
was common, it was GPL, etc).
Because bash was essential, packages started using it. It was
guaranteed to be on the system, and hence rapidly became a defacto
standard.
Now, because packages and people took to heart the implied
promise of things labeled essential, we can't easily do away with it,
without breaking expectations and systems.
The work and expense and the instability involved demand a
darned good reason.
I am surprised that I have to spell this out.
Santiago> Do you mean it is ok that we do something "just because we
Santiago> do it"? Frankly, I usually feel better when I do something
Santiago> because there is a real reason to do it.
Do you understand now? Do I have to go into more excruciating
detail?
Santiago> I don't buy the idea of an "implied promise". Other
Santiago> essential packages have been downgraded to required, and
Santiago> nobody *ever* talked about "implied promises". Yes, bash is
Santiago> *currently* more important than those packages, but the
Santiago> difference here is only cuantitative, not qualitative.
Please refresh my memory. The only package so downgreaded, in
case you have forgotten, was made into a virtual package that also
was a dependency of an essential package; and the reasons for doing
so were that other packages could provide it.
Does not look like no one raised a fuss. To all intents and
practices, an awk package is essential.
You make bash a virtual package, and make an essential package
depend on it, and all packages that provide the virtual bash can be
invoked as bash and not break any scripts; and no one shall talk
about implied promises either.
Santiago> Well, if this is really your opinion, I would call it a
Santiago> fundamentalism.
Sure, go ahead, call my opinion names. Have you lost the
debate already? I think you are far too eager to mould the system to
your way of thinking, with no though to the expectations and systems
that change might break. I object to that high handed manner of
operation.
Santiago> Raul has explained very well which would be the right way
Santiago> to do it, do you have any objection to any of his points?
Santiago> (Note: We are talking for the time another posix shell is
Santiago> made available, not before).
Yes, I do. Raul has talked about how to do it if we are to do
it. I see no compulsive reasons to do it, so far.
Santiago> This is a really really bad example. As far as I know,
Santiago> there is not a replacement for ldso, and every dynamically
Santiago> linked ELF binary needs it. Compare this with the fact
Santiago> that already 90% of our shell scripts are already /bin/sh
Santiago> and not /bin/bash and you will see how bad your comparison
Santiago> is.
Who the hell is talking about not having interchangeable
/bin/sh? I am talking about making bash onn-essential, and
potentially not present on the system You are confused.
There is no replacement for /bin/bash.
Santiago> But this is the circular argument again!
Santiago> Once bash is non-essential, we would not risk the system at all by
Santiago> removing it since it would not be essential anymore!
You are too late. I do not think bash can be made
non-essential easily. And I see no reason for doing so.
manoj
--
Politicians should read science fiction, not westerns and detective
stories. Arthur C. Clarke
Manoj Srivastava <srivasta@acm.org> <http://www.datasync.com/%7Esrivasta/>
Key C7261095 fingerprint = CB D9 F4 12 68 07 E4 05 CC 2D 27 12 1D F5 E8 6E
--
To UNSUBSCRIBE, email to debian-devel-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: