Limits to what can be done without source
Andy Green
andy at warmcat.com
Wed May 17 19:16:22 UTC 2006
Les Mikesell wrote:
> On Wed, 2006-05-17 at 02:31, Andy Green wrote:
>
>> Consider for a moment that you are Mike Harris and this complaint lands
>> on your monitor. Say you want to help. What exactly is your next move?
>> Look at the ATI source to see what has gone wrong? There isn't any,
>> ATI refuse to give it out.
>
> Did it work in a previous fedora release? Does it work with
> other distributions? In other words, did you change the
> interface in a way responsible for the breakage?
ABI churn is not the only problem with binary blobs. Point in case I
saw on this list in the last couple of days, Adobe Acrobat blew chunks
on a double free. This is not an ABI problem but a hidden bug in the
binary blob.
>> Should you, as valuable resource, then bust
>> out a debugger and spend miserable weeks trawling through disassemblies
>> to try to understand the origin of the problem, likely caused by an ATI
>> programmer that is sitting on the source that could help you solve it in
>> hours rather than weeks? What exactly can you do to help other than
>> shrug your shoulders and tell you to complain at the guys sitting on the
>> sources?
>
> Provide a documented and unchanging interface so if something works
> today it will still work next week.
That does not follow for the same reason... a stable ABI would be nice
but that's not what one can expect with Linux. It won't guarantee
binary blobs becoming paragons of coding virtue and to provide immortal
functionality either.
I have to maintain a chunk of kernelside code and tracking the whiplash
on the kernel can be a PITA, so your point is understood. But it's not
like the only issue with opaque binaries is that the ABI keeps changing.
-Andy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 4492 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://listman.redhat.com/archives/fedora-list/attachments/20060517/984cb058/attachment-0001.bin>
More information about the fedora-list
mailing list