Re: Regarding LSR translation work

[Graham Percival]

On Sat, Apr 21, 2012 at 11:33:49AM +0100, James wrote:
> Err.. not sure what you mean by that technically but it's complicated
> enough with snippets as it is, to follow the translation method adds
> even more complexity (if the CG is anything to go by).

+1

> We want to encourage contributors not push them away.

Arguably the complexity of the current system pushes translators
away, so Jean-Charles' has a point there.  However, I am extremely
skeptical of any major work on this part of our build system and
policies.  Unless you know that you know more about the build
system and LSR than I do[1], I really don't recommend getting into
this.  I see any work on this becoming a huge time sink which
probably won't end up with a simpler system anyway.

[1] John Mandereau is the chief candidate for this.

> >From a very basic-no-progamming skill perspective, why can't we just
> have an extra texidoc entry in the snippet itself and add the
> translation manually, like we would for any updated snippet?

Because then it would be overwritten whenever we do a LSR import.

> In fact, why do we even need a texidoc string *inside* the snippet? I
> don't use one for @lilypond examples.

Because snippets come from LSR, and that needs a texidoc to
display some text for the snippet.


I really don't see any way to rescue the snippet system.  It was a
gamble that hasn't paid off; it was supposed to be a standalone
system that would let normal unprivileged users contribute to the
docs with almost no oversight by anybody with git access.  With
all the developer effort that's gone into build and maintaing it,
we could have made it easier to do stuff in git and/or trained
more doc editors.

I view any major effort on the snippets as "throwing good money
after bad".  Or, to put it more technically, it's reached the
level of the sunk cost fallacy.

- Graham