TT brainfart

[Dave Hodgkinson]

On 3 Jan 2007, at 15:53, Mark Blackman wrote:

> Dave Hodgkinson wrote:
>> On 3 Jan 2007, at 15:18, Mark Blackman wrote:
>>> Dave Hodgkinson wrote:
>>>> How easy or even useful would it be to have the TT stash stashed  
>>>> in memcache?
>>>
>>> easy, but would probably make more sense to cache the result
>>> of 'process' rather than its inputs, i reckon.
>> Dependency graph hell.
>
> I suppose it does depend on how complex the templates are
> and why you want that degree of caching. I probably missed
> the point of the query anyway.
>
> possibly, not very usefully, I was thinking about the simplest  
> case, same simple template used repeatedly with scalars, lists,  
> hashes and
> nested versions thereof.

And I didn't give the reply the consideration it deserved.

My understanding of the stash is that it's a store of compiled but as- 
yet-
un-interpolated templates. This is a good, quick win IMHO.

Once data is in the template, all bets are off as we need to know  
about the
timeliness of the source data. See the inability of SSI's even to  
generate
correct Last-Modified headers.

Anything further down the pipe requires domain knowledge, ie: cache  
navigation INCLUDE,
cache page content but don't cache the box with the user's cart in it.

-- 
Dave Hodgkinson - Music photography
http://www.davehodgkinson.com/