cURL
Haxx ad
libcurl

curl's project page on SourceForge.net

Sponsors:
Haxx

cURL > Mailing List > Monthly Index > Single Mail

curlpp mailing list Archives

Re: [cURLpp] Need for issue tracking?

From: Piotr Dobrogost <curlpp_at_2008.autoera.pl>
Date: Tue, 25 Nov 2008 02:28:18 +0100

Brad

Brad Hubbard wrote:
>
> Piotr Dobrogost wrote:
>> The thing is what for are we building curlpp as a library if only a very small percentage of code goes into the library object file? Maybe we should abandon building curlpp as a library and just tell users to take a source code and compile it with their projects. If not, and if we're going to still support building curlpp as a library we should at least give users a chance to make it a real library with all code inside and not just a few functions that happened to be non templated.
>>
>> I can summarize this problem in this question
>> "What are the benefits of building a library file out of curlpp if most part of functionality used by library's users would end up being generated from source and put in users' executable?
>> The whole idea of building libraries is to allow many programs which need the same functions to use ones placed in a library's file. In case of curlpp any program that uses it generates its own versions of functions and makes almost no use of the library's file. That's against the idea of libraries. It's like someone would write smart pointer template library with extra bonus let's say function giving current system time. And the author would say "Here you have a library. You have to build it and then you can use it." In fact what would be built would be only funny bonus function and nothing from the real functionality. So for all users that would be easier not to bother with building it at all.
>>
> What about boost that is pretty much all headers with relatively little
> in the library files. Why did they make that design decision do you propose?

I'm not sure I understand your question. If you are asking why most of boost is delivered as source code only then the answer is - they don't have a choice. Boost libraries are written in such a way as to be able to operate on user defined types. That's the real power of templates. They can't compile their libraries because it can be done only if the library knows user's types. So only user of the library can compile it _together_ with his code.

In case of curlpp situation is different. Curlpp doesn't operate on user defined types and uses templates only "internally". As such it can be compiled into library object file.

Regards
Piotr Dobrogost
_______________________________________________
cURLpp mailing list
cURLpp_at_rrette.com
http://www.rrette.com/mailman/listinfo/curlpp
Received on 2008-11-25

These mail archives are generated by hypermail.

donate! Page updated November 12, 2010.
web site info

File upload with ASP.NET