[pLog-general] It's about time... pLog 1.0 is out :-)

[Oscar Renalias]

On 1 Apr 2005, at 22:29, Benjamin Krause wrote:

>
>> yeah it's a the presentation level... but if you look at for example 
>> DefaultAction, you'll see that the first thing we do there is ask 
>> whether our view is cached. If it is, then we don't even bother 
>> loading the data and we go ahead with what was cached so we are 
>> saving a lot of processing time too.
>
> okay i see your caching strategy.. and you're right, there is not that 
> much we can do differently.. this will work with a small number of 
> blogs, but you already stated in the code that there is optimisation 
> potential :)
>
> i would have suggested to add a cache to the data layer, e.g. on an 
> article to cache the text, body and comments.. and if you call 
> setText() or whatever, to throw away the cache..
>
> this might grant an higher performance on a high trafic page. and yeah 
> of course you dont dump those files to disk, but to a shared memory 
> section.
>
> i guess we have 2 ways... just optimize the way you clean up the 
> cache..  but i don't see a better level then to throw away single 
> pages (instead of all pages of a blog). this will still result in a 
> lot of sql requests. but most of the work is done already ..

I like the idea, though it make take some time until we implement it 
and iron out all the kinks... But I'm all for it :-)

Regarding the template caches, I would also love to have a more 
fine-grained cache system where only the pages that are not needed 
anymore can be thrown away, but it is not too easy, namely because we 
don't know what appears where (for example posts in the "recently..." 
section appear everywhere because of header.template/footer.template so 
updating one of those is forcing us to throw away everything -- and so 
on)

> on the other hand we could try to implement some sort of data layer 
> cache, where we will cache every single bit of information and just 
> throw away caches of really changed information. but you're right, on 
> a not-optimized webserver, where you dont have a /dev/shm or something 
> like that, this might even slow the whole thing down ..
>
> i guess it depends on the target group of your software :) or we write 
> both cache mechanisms as modules..

a data cache should be optional, like the template cache is. Those who 
need it can enable it and those who don't need it will live without. I 
suppose that a well designed cache system will allow us to easily do 
this anyway... :-)

Oscar