Re: [PATCH] mime-type detection "improvement"

mark benedetto king wrote:
> On Wed, Jul 09, 2003 at 09:05:07PM +0200, SLOGEN wrote:
>
>>- 1233 parses mime.types file for every time a type-guess is req'ed, I
>>do one parse into an apr_hash_t, and use that for lookup

> -1 on this approach. AFAIK, there is an engraved-in-stone policy against
> non-const static data. The approach used for other configuration (load
> it, associate a handle with it, and store it in the client-context)
> would be an acceptable alternative.

Thanks for your pointers on that. I didn't really know what policy SVN
had on data that needs non-scoped storing (for dynamic programming,
caching ...).

Since the mime-type data is bound to the lifetime of the client, the
proper place to store it is of course in a pool which lasts for the
client life-time.

I'll browse through the configuration implementation to find the proper
idiom for storing the hash-data then.

-- 
Helge
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org