Re: STL revisited

On Sun, Sep 03, 2006, Robert Collins wrote:
> I'd like to raise the STL again as an option for squid3 rather than
> NIH'ing a whole bunch of core infrastructure.

Its cool. The STL looks nice. The really trippy thing will be:

* will it affect performance?
* how much will it cut the current codebase down by?
* if they're at odds; which is more important?

And:

* can it wait for after -3? :)

My suggestion is to wait until -3 is released, then create an "STL" SF
branch and nut out the code to incorporate STL. We can then compare
codebase clarity, performance and portability.

Adrian

> The STL, and boost:: - the STL's 'incoming pipeline' provides a tonne of
> really useful tools for efficient data implementations of trees, hash's
> etc etc etc.
>
> Last time the big concern was over readable compiler errors.
>
> I dont think that has improved hugely over the last 2 years, but I think
> that its a net win - there is usually a clear reason for the error if
> you read at the top, and the amount of code that we could delete, or
> stop writing, is huge.
>
> -Rob
> --
> GPG key available at: <http://www.robertcollins.net/keys.txt>.
Received on Sun Sep 03 2006 - 10:10:43 MDT