Re: [9fans] rc vs sh

[tlaronde@polynum.com]

[Since the previous one did not reach the list (?), I send it once more]

On Tue, Aug 28, 2012 at 07:12:15PM +0300, Aram H?v?rneanu wrote:
> > Yes. This is what is done by the R.I.S.K. framework used for building
> > KerGIS and kerTeX.
>
> I'm pretty sure that R.I.S.K has more than 2,250 lines of code. That's
> the LOC count of \.(ba)?sh$ stuff in the Go tree. Also, nobody seemed
> to mention that Go also ships with rc files to build on Plan 9...


result of "wc -l rk*":

     838 rkbuild
    1121 rkconfig
      60 rkguess
     247 rkinstall
     256 rkpkg
    2522 total

This is with comments of course. The rest are the trivial parameters
files for each system. (rkguess is used to sketch such a parameters file
on a new system.)

And this does what no other framework does: be able to remove
intermediary products, such that you can compile a resulting n
megabytes package with just slightly more than n megabytes of
space...

To be clear: I'm answering the "there is not anything existing proving
this can be done differently". I'm not arguing about the choices
of the Go developers:  they do the work; they do as they see fit.
Period.

And as for KerGIS vs. GRASS; kerTeX vs. TexLive; it costed me less time
to do R.I.S.K. from scratch, knowing thus anything about how it works,
how to fix, how to improve, why it fails, than to try to use existing
monsters.

Sending to /dev/null is the developer's primary tool. (As well as for
the readers of my messages, probably...)

--
        Thierry Laronde <tlaronde +AT+ polynum +dot+ com>
                      http://www.kergis.com/
Key fingerprint = 0FF7 E906 FBAF FE95 FD89  250D 52B1 AE95 6006 F40C