Re: bug #21319 - errors in separate_translation mechanism as of 070618-1

[Dmitry Borodaenko]

Hi Boud,

Thanks for all the details, it really helped to clear things up!

On Tue, Feb 19, 2008 at 2:38 AM, boud <address@hidden> wrote:
>  > The only reason I can imagine to have both a bug and a patch open is
>  > if the patch fixes several different existing bugs with the same code
>  > (as in, fixes can't be isolated). Why open a bug that merely points to
>  > an existing patch?
>  My worry was that since there was nothing listed in the list of bugs,
>  people might miss it, even though it's a set of (related) bugs.

I'm not yet convinced that this is a set of bugs, so this point is
moot. One thing that is definitely missing from the bug description
is, well, a bug description that would allow reader to at least
understand whether his own issue is related to the bug in question.

Also, please consider that overobundance of information turns
information into noise. I am developer and I have a hard time wading
through all Samizdat bugs and patches on Savannah, I imagine some
users might drown threre.

>  i agree that code duplication is against the whole spirit of ruby

Not just Ruby, code duplication is a major sin in software development
in general. It makes code harder to understand (not only by forcing
you to read more code, but also by obscuring the fact that different
parts of code all do the same thing) and harder to modify (by
increasing the risk that changes to one duplicated spot won't be made
to the others, or that changes blindly copied from one place will not
do the right thing in the other place). It is also the first "code
smell" (see "Refactoring" by Martin Fowler), a sign that design of the
program doesn't fully match its purpose.

>  On the other hand, the problem
>  is important and IMHO is a quite high priority for being fixed.

Not high enough to accept this patch as it is.

If I don't find a way to improve handling of translations with minimal
impact on the Samizdat code, this patch will have to wait until after
0.6.1: current version as it is has enough benefits over 0.6.0 even
without it.

>  > The patch also appears to fix several unrelated issues in one go.
>
>  This i don't see, except for the comment on #20303, but that's just in
>  the latest comment.

That's one of the things I meant here.

>  IMHO most of
>  them are part of the same issue: the user clicks (or cannot click due
>  to missing button(s)) and expects something dealing with the message
>  that s/he saw, but instead gets an equivalent message (source or
>  rendered) but in another language (if the initial translation was
>  correctly done).

One other thing I meant here was that I didn't find a short
description like the above that would explain what all the different
cases have in common. The above clarifies that a bit, although it's
still not good enough: "user expects" is a subjective criteria, one of
the reasons I don't understand this problem might be that my
expectations may be different from yours.

>  OK - i think my error in terms of keeping things neat was that i tried
>  updating the "original" separate translation patch whereas i should have
>  started a new entry - probably as a bug with suggested fix?

As long as it was a bug in released code and not in the patch in
question, yes, you should have created a new bug.

Hm, now that reminds me. At first I implemented my own version of
separate translations that was different from your patch (see
ChangeLog for 2007-06-18), and then I reverted back a major chunk of
these changes (2007-11-04). If the last version you have seen is
20070818, that could be one more reason why I don't understand you.
Current model of handling translations is different in at least one
important way: in the "full" mode, message is not replaced with a
preferred translation.

>  i'm not sure where i should put this. At least putting it here
>  on the mailing list can't hurt, since the issue has been raised and
>  IMHO is important. i guess that the neatest thing, if we don't solve
>  this soon, is to add this to bug 21319 and close up patch 5882.

Good plan. However, please make sure you try out the latest snapshot
first, I suspect this problem is gone or at least mitigated since
2007-11-04.

-- 
Dmitry Borodaenko