a proposal regarding libraries - loading libs from web on demand?

Isto Aho isto.aho at dnainternet.net
Fri Mar 3 13:44:03 EST 2006


Josef Svenningsson wrote:

> On 3/3/06, *Henning Thielemann* <lemming at henning-thielemann.de
> <mailto:lemming at henning-thielemann.de>>

>     On Thu, 2 Mar 2006, Josef Svenningsson wrote:
>     > They can always be improved upon later.
>
>     How can you improve the API later without breaking code that
>     relies on it? 
>
>
> You can't. You have to brake the code. Which is not that big a deal.
> The old version of the library doesn't disappear off the face of the
> earth just because there is a new version. I agree that backwards


This was the reason for the thought about the division of the libs clearly
to the mature ones and the developing ones.  If this was shown somehow
to the lib users, at least they would know, if they take a larger risk that
their code will break in near future -- I have to admit that this doesn't
sound that good any more...

There are some systems (could be called interpreters) where the new
versions of the libs can be loaded from the web in the console
(something like ghci).   What would it require to have a repository of
old versions of the libs, so that no code would ever break, if the
compiling machine had internet connections?  Something like
        import Directory version 4.7.3
and somewhere the address of the repository?
The import line could have it at the end overriding the default setting...
Does the cabal do that? (Sorry, knows nothing about cabal at the
moment.)

br, Isto




More information about the Libraries mailing list