Re: midi2ly doesn't follow time signatures (beat divisions)

[David Kastrup]

Urs Liska <address@hidden> writes:

> Am 20.02.2014 03:09, schrieb Paul Morris:
>> For what it's worth, if anyone is looking for ways to improve
>> midi2ly, the first footnote of this blog post:
>> http://blog.steinberg.net/2014/02/development-diary-part-the-fifth/#fn-523-1
>>
>> notes that midi2ly does not follow the beat divisions of the time
>> signature when it decides how to represent tied notes (or rests):
>
> Yes, the very best response to such a post would be to comment on it
> in a few weeks reporting that we have fixed the issue ;-)

Indeed.  We don't have any people working dedicately on midi2ly
currently I think.

> But the rest of the post is also interesting. You surely noticed that
> Daniel states the obvious deficiencies of Finale and Sibelius with
> rest positioning, displays LilyPond's output as considerably better
> and states that they want to be even better than that.

It's easy to say "good luck with that", but there is an underlying
lesson in there: it's probably not easily doable to cast the rules of
good engraving into actual code with either Finale, Sibelius, _or_
LilyPond.  With Sibelius, their corporate parent apparently decided to
stop investing any more manpower in the improvement of typesetting and
gave the core developers the boot.  With Finale, one has the impression
that they have not made quality of typesetting a high priority anyway.

With LilyPond, we are working with volunteers and enthusiasts, and they
will often put in more work than one could justify to management for a
given result.  We are reaching the point of diminuishing and
self-defeating returns often already: improvements are hard to do, and
get in the way of doing further improvements.

So what is called for is an architecture that makes it rather
straightforward to specify good rules of typesetting rather than
spelling out code that conflicts with other code in ways that are an
artifact of the coding rather than based on a specification.

Starting from scratch, of course, provides a chance to design an
architecture suitable for the full requirements one is familiar with
rather than evolve an architecture while the requirements become more
and more apparent.

Or putting it more concisely: stealing good ideas requires good code.
You don't use rickshaws for a copper heist.

> I think this is a demonstration of respect (that is noticeable in most
> of what he writes anyway).
>
> So please don't start ratning in the comments (again). We had this at
> the beginning of its existence, and it's good that we've overcome the
> urge to abuse Steinberg's blog as a free "advertising platform".

Yup.

-- 
David Kastrup