Argument list too long.

Mike A. Harris mharris at redhat.com
Sat Sep 13 08:19:20 UTC 2003


On Thu, 11 Sep 2003, Dag Wieers wrote:

>> > Ok, then I'm on the right list afterall ;)
>> > 
>> > Please Red Hat change rpm so that I can sign packages without giving them 
>> > as arguments and without having to enter my passphrase for each (set of) 
>> > packages.
>> > 
>> > This seems really silly to me however and this was just an example. Fact 
>> > is that most scripts suffer from this and as usual people will only notice 
>> > this when the script fails (with potential data-loss, manual correction 
>> > and other horrible things). </scare-tactics>
>> > 
>> 
>>      But this is _exactly_ the point that goes against your own argument.   If 
>> you accept that there has to be some upper limit then a properly written
>> script will always have to guard against such buffer overflows.   No matter
>> what limit you pick.
>
>I'm not arguing that if you increase the limit, there's no limit anymore. 
>(Doh!) The only thing I'm trying to say is that maybe after 10 years (or 
>when was this 128Kb limit added?) memory isn't much a problem now and for 
>the common case 128Kb could easily be 256Kb without anyone noticing (and 
>without causing any more problems).

There's one definite immediate problem it would cause.  If it did 
actually work for someone, then their script would break on 
systems that didn't have 256Kb, and the problem is worse.


>And as you say most people wouldn't have this problem, well,
>increasing it to 256Kb will not cause any extra problems either.
>It's not that you are obliged to fill the 256Kb or anything. No
>overhead.
>
>So, since you haven't had any problems with it, why are you against it ?

Probably because it is the wrong solution to a non-general-case 
rare problem, that has ramifications that affect everyone, and 
break compatibilty with virtually every piece of software out 
there.

-- 
Mike A. Harris     ftp://people.redhat.com/mharris
OS Systems Engineer - XFree86 maintainer - Red Hat





More information about the devel mailing list