On Thu, Mar 11, 2010 at 2:16 AM, Gary Maxwell <[email protected]> wrote:
> Olaf van der Spek intoned on Wednesday, March 10, 2010 4:10 PM:
>>
>> Yes, I understand your case. Although in that case the code wouldn't
>> change much. You'd just disable that header instead of enable it in
>> certain cases.
>>
>
> Are you asking people to change existing, stable code which works fine right 
> now? That may be fine if the original developer is still around, not so fine 
> if the remaining developers treat libcurl and the code which accesses it as a 
> black box.

No, I'm asking people to fix their server configs.

>> So because some servers don't have proper logic to decide on
>> compression, we should disable compression in the client by default?
>>
>
> Interesting you should pose this as a rhetorical question when compression is 
> already disabled by default.
>
> Some clients (mine) do not use compression, and will never use compression 
> due to memory footprint issues and small size (1K ~ 8K) of the datasets 
> downloaded with libcurl.

To me, it seems perfectly reasonable to compress 8 kb (assuming it can
be significantly compressed).

>> Why?
>> Instead of code to enable it in some cases, you have code to disable
>> it in the other cases.
>>
>
> It adds complexity for existing libcurl users who will be forced to modify 
> their libcurl application when they upgrade the library, which in turn adds 
> additional regression testing.

Only when the server's config is wrong.

> It adds complexity when applications suddenly break when the libcurl library 
> is silently upgraded as part of an automatic update for servers and desktops.

Again, IMO, that's a server-side bug. If apps fail due to this, it
might be a good idea to fix stuff regardless.

Olaf
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html

Reply via email to