On 11 June 2010 15:59, Colin Strickland <[email protected]> wrote:
> As ever, there's no such thing as a magic answer. I don't disagree with 
> anything you say, but you don't always have that luxury of choice.
>
> Sometimes, when you are programming for the web in the large, execution time 
> is considerably significant. Admittedly, most people don't work at this 
> scale. I myself
> have worked on some moderately busy web apps before, and consequently spent 
> much time doing ugly computational things I disliked inside the presentation 
> layer.
>
> You need to optimise in the right areas. Look at Facebook and their bazillion 
> lines of PHP templating and native PHP compiler. Look also at Twitter, with 
> their
> beautiful MVC RoR roots, and their recurrent whale pictures.

I've worked on some pretty busy sites, but none busy enough to reach
the point where the page templating speed of Mason or TT had any
impact (unlike IMDB where even relatively little costs per request can
add up to make a noticable difference). OTOH I've seen plenty of
handrolled inhouse templating systems totally fail to scale or
suddenly degrade badly in performace when core perl fixes or changes
regular expression implementation details that were being relied on by
the original programmer.

It's all about bottlenecks and designing for scaling (rather than
performance), if you have a sensible caching strategy and pre-build
snippets, etc then TT can be blindingly fast, and still provide
awesome benefits in terms of ease of development.

The main thing to remember is that technology can't (usually) solve
bad design decisions, good design is considerably more likely to scale
and perform well, and be easier to improve/tune/pimp when you need to.

A.

-- 
Aaron J Trevena, BSc Hons
http://www.aarontrevena.co.uk
LAMP System Integration, Development and Consulting
_______________________________________________
BristolBathPM mailing list
[email protected]
http://mailman.bristolbath.org/mailman/listinfo/bristolbathpm

Reply via email to