[wp-hackers] GSoC Proposal: Caching in core

[Andy Skelton]

On Tue, Apr 1, 2008 at 2:10 PM, Mahmoud Al-Qudsi
<computerguru at neosmart.net> wrote:
>  Maybe you could explain the other benefits of a tagging-based system vs. the current implementation?

The current implementation has a delete function for deleting
individual items and a flush function for flushing the entire cache.
It does not provide any method for flushing groups, though this would
be a trivial patch. Even so, groups are a one-to-many structure.

The versioned tag-based cache enables many-to-many grouping and
instant mass extinction of related items. There is no need to scan an
index to find the related keys because simply incrementing the tag
version makes all related items unreachable.

This does not translate directly into better performance for
WordPress. All it does is provide fine-grained control over related
items in a persistent cache.

I do not support switching WordPress to a versioned tag-based cache.
Extensive modifications would be required to see any benefit from
that. I hope the concept can be proven without that monumental effort.

The GSOC proposal from Thomas is huge but informed and inspired. I
expect it to trim down after an examination of goals vs. time.

Andy