Re: svn commit: r1613854 - in /subversion/branches/authzperf: BRANCH-README subversion/libsvn_repos/authz.c

On 28.07.2014 12:58, Ivan Zhakov wrote:
> On 28 July 2014 14:52, Branko Čibej <brane_at_wandisco.com> wrote:
>> On 28.07.2014 12:40, Ivan Zhakov wrote:
>>
>>
>>
>> Anyone who want to improve authz performance should consider to
>> separate config file parser and config storage. To read authz
>> configuration in appropriate custom structure directly and avoid read
>> authz file in intermediate svn_config_t.
>>
>>
>> That's on my todo list, as the very last step. The parsing overhead is the
>> least of our worries right now.
>>
> Great! I just wanted to share my ideas on this topic. As far I
> remember parsing overhead may be significant for big authz files.

I don't doubt that. My point is just that there are likely to be far
more expensive lookups than expensive parsings; and parsing is trivial
compared to optimizing lookups, so getting the hard part done first is
just common sense.

-- Brane

-- 
Branko Čibej | Director of Subversion
WANdisco | Realising the impossibilities of Big Data
e. brane_at_wandisco.com