Warp wrote:
> 
> Alessandro Coppo <a.c### [at] iolit> wrote:
> : anything UNIX related has .so's
> 
>   But there are several computers out there which use unix. An .so file for
> one system can't be used in another. There should be pre-compiled dynamic
> libraries for Windows, PC Linux, Sparc Linux, Alpha Linux, Mac Linux, FreeBSD,
> NetBSD, OpenBSD, MacOS (many different versions), Sparc Solaris, PC Solaris,
> Digital Unix, OS/2... The list is endless.
> 
>   And what's the point anyways? I really don't see such a big advantage of
> this. It just introduces really big maintenance and incompatibility problems.
> 
First, I must say that I agree with Warp.
But the distribution might be on source, not binary.
The only trouble is that the syntax sugar to generate a .so differs
greatly from one unix to another. (Including between version of the same
unix-provider). And then every user must get access to a C++ compiler, 
instead of using directly. 
Moreover, it is usually required that the .so to be in a special place which
required root priviledge to install...
I really feel this approach should be avoided.

Maybe what really need the first requester was a simpler way to add 
object/pattern/.. so that instead of having to always patch the same 
common file (parse.*, tokenize.*, whatever) and then a set of separate
files for the new thing, all could be simplified to just one single file
to patch (with only a simple #include ... ?) and the new set of files 
for the new thing.

IMNSHO, We are not ready for 4.0 yet. Let's play first with 3.5, then having
some super/mega/giga/hyperpatch. Then maybe, some people will start thinging
how to waste more CPU cycles with a 4.0 and a complete redesign.
But probably not before 3 or 5 years.

Post a reply to this message