On 03/07/2015 11:27 PM, Alexandre Torres Porres wrote:
But if you load kalashnikov first, well, now you won't get cyclone's version when you type "uzi", but kalashnikov...
so, well, kinda confusing...
yes, very confusing. but not something unheard of, and i think humans are quite good at dealing with such things. imagine, one of the organizers of Pd~con 2007 is called "Alexandre" and one of the organizers of Pd~con 2009 is called "Alexandre". one is inclined to think that they are the same (after all, they do similar things and go by the same name), but it turns out that in fact they are totally incompatible (crashing *your* place in Montreal might make me end up sleeping on the street!)¹
I'd vouch for trying and eliminating the redundancy and equal names somehow. I'd suggest killing the alias name of kalashnikov, it'd solve all that for Pd Extended.
and get rid of cyclone's [uzi], as it already has [Uzi].
but again, useless and pointless discussion if we're not dealing with an update of Pd Extended right now.
no, i think that the discussion is important, as it shows one of the big problems with the architecture of a monolithic Pd-extended.
the question is: do "we" (the hypothetical PdX maintainers) provide a consistent system where everything is nice and easy; or do we just provide a largish collection of libraries for all kind of problems.
i think the 1st option is *totally* out of scope.
the fact is, that PdX currently *is* a largish collection of libraries, sharing a significant overlap (both in functionality and in naming).
it would require multiple fulltime jobs to sort this pile into consistent stack (and it would take a similar number of workpower to keep it in that state!). and once you have eliminated all redundancies, what you have gained is a centralized distribution of a decentralized development process that has broken any old patch by discarding backwards compatibility.
just *having* such a distribution does not mean that anybody will use it (e.g. those people that do not upgrade from PdX-0.42 to PdX-0.43 because...) nor that anybody will *develop* components (externals,...) for it (apart from those fulltime jobbers).
my point has always been that we should *embrace* the multitude in Pd, rather than eliminate it.
there is no harm in having "uzi" and "uzi" and "uzi", as long as it is clear which one is currently used (something Pd still lacks, despite hans' great work in the past).
having said all that, Pd-l2ork probably already does a decent job in providing a consistent distribution (but i haven't checked recently; and of course, l2ork/dsis also *added* a few new objects the functionality of which is already included in PdX - so not exactly minimizing the pool of objects either) - most likely because it *is* powered by institutional backing (see "fulltime jobber").
¹ actually i don't know how often you and alexandre have been confused; i know for sure that a lot of people mistake me for hans-christoph although the name is really not *that* similar, and we often shared very differing opinions. do you have any suggestions for my case :-)?