Re: Some more plane hacking...

On Sat, 18 Aug 2001, Tomas V.V.Cox wrote: > Stig Sæther Bakken wrote: > > > > > > I buy all your arguments about maintainability etc. But there are > > some things that either _must_ be done in C (efficient asynchronous > > IO, low-level database abstractions, GTK bindings or whatever) or that > > would have significant performance benefits for everyone (the PEAR and > > PEAR_Error classes will be used all over the place, why not save a > > file parse?) > > I buy also your "speed improvements" ideas, but I'll try to give an idea > about the problems with core C extensions. One day someone wake up with > a new cool ultraRaiseError method and implemented it. Of course, Pear > components starts to use it all arround. From this moment I must to stop > upgrade all pear components I use in production envs, because I can not > spend so much time compiling and testing. The life cycle of a php lib > release is much more short than the one from a C extension. > So? What if I implement a super cool php function that all the pear components start to use? guess what, you have to upgrade PHP in that instance. If it makes more sense to implement it in C, then it should be done in C. PEAR has always had the intent of implementing certain extensions in C (or even *gasp* C++). Scripts you right for PHP 4.0.6 won't work with PHP 4.0 if you use the new features of PHP 4.0.6. -Sterling