develooper Front page | perl.bootstrap | Postings from July 2000

Re: Working Group Proposal

From:
Nick Ing-Simmons
Date:
July 24, 2000 13:41
Subject:
Re: Working Group Proposal
Message ID:
200007242040.VAA20578@gabrielle.tiuk.ti.com
Horos <horos@earth.he.net> writes:
>> >Sure,
>> >it might make _p_erl faster and leaner, 
>> 
>> Actually it probably would not make that much difference to '_p_erl'.
>> 
>
>How about in the designing/creation of a compiler? 
>How about in creation of 
>IDEs with syntax coloring && intellisense types of things? Some may not care 
>for them, but large groups of people do...

Which would all benefit from having a formal grammar - but the question 
was how it would affect 'perl' (the program) - and my answer was "not much".

>
>Ok, I guess my concern is being able to embed perl inside of other applications,
>and the memory cost of doing so. ( ie: the scalability problem ).

The grammar constructs that are allowed have little impact on that - it is 
the code behind the operators etc. that costs.

>
>And yet ANOTHER concern is the existence of various holdovers from perl4, stuff
>like getservbyname which really has no right to be in the core proper. 
>( ie: the legacy problem ).

Right - that is what is different about perl6 - Larry has said we can 
rationalize those this time where he did not allow himself to do that 
with perl5.

>
>And, on the other side of the coin, another concern about perl5-porters 
>was the inability to add stuff to the core without going through a huge tug of 
>war. ( ie: the immutability problem )

Right again - so this time we try even harder to make modular additions 
possible - perl5 made a major jump in this direction but there is scope 
for more.


>
>And finally, the fact that perl didn't natively support certain protocols 
>(http comes to mind) and needed external programs to bootstrap itself (lynx,
>ftp, etc inside of CPAN) ( ie: the portability problem ).

Which is the inverse of the above - on the one hand the "core" should 
not have getservbyname() let alone HTTP/FTP "built in" - on the other 
the "distribution" should be self contained. 

>
>I think we need a method to make perl both really small and large at the same 
>time.   I'm open to suggestions, but I think that being big and monolithic 
>has a lot of drawbacks if perl is to spread.. Linux does it right, I think, by 
>allowing modules to be imported/exported from the kernel at will.

Which is what all the "in the core" debates were about - do it as a module
not in the core. Then we have debates about which modules are bundled.

-- 
Nick Ing-Simmons




nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About