Re: [SPAM] Re: GSoC 2014

[Janek Warchoł]

2014-02-19 23:21 GMT+01:00 David Kastrup <address@hidden>:
> Janek Warchoł <address@hidden> writes:
>
>> 2014-02-19 18:41 GMT+01:00 David Kastrup <address@hidden>:
>>> Urs Liska <address@hidden> writes:
>>>> Janek's Tie Crusade?
>>>
>>> There is a three-month limit as far as I can see.  There seems little
>>> point in proposing stuff that a student cannot realistically be expected
>>> to wrap his head around in the given time frame.
>>> [...] With something vague and complicated
>>> like "tie crusade" or "slur formatting", I don't see anything happening
>>> with a tangible result.
>>
>> I estimate the time needed for fixing tie formatting to be 120-200
>> hours (for someone roughly familiar with LilyPond codebase).
>
> What does "fixing tie formatting" mean?  That no tie needs tweaking any
> more?

That all of the example ties from my research (a couple hundred ties,
they cover almost all real-life tie occurrences) are formatted
correctly by default.  Assuming my examples are representative (and
believe me, i know what i'm doing) this should mean that 99% (maybe up
to 99.9%) of all ties in Lilypond-made scores will be good by default.
 (Currently only about 64% ties are correct, as real-life statistics
has shown http://lilypondblog.org/2014/01/engraving-statistics-slurs-and-ties/).

>> When we take into account that such estimates usually are 50% off, and
>> add 100 hours for getting familiar with the codebase,
>
> 100 hours for getting familiar with the codebase.  To get to a state
> where you can sensibly design a new tie formatting strategy that fits
> into the current code and interacts with it.

As i said, i already have a pretty good knowledge of this area, and
i'd be willing to mentor such a student.

>> 2014-02-19 20:16 GMT+01:00 David Kastrup <address@hidden>:
>>> Urs Liska <address@hidden> writes:
>>>> That may well be true, and my suggestion wasn't too educated.
>>>> But I had the impression that at least the Tie project was of a
>>>> reasonable size,
>>>
>>> What _is_ the "tie project"?  There are lists of ugly ties.  That's not
>>> a project.
>>>
>>>> and equipped with enough pre-research.
>>>
>>> I don't see anything that would serve as either a direction or a
>>> starting point.
>>
>> You know what?  It's VERY frustrating to spend 200+ hours (during the
>> last 2 years) collecting examples and researching tie formatting, and
>> then see someone - especially an involved developer like you - say
>> "what tie project? what research? i don't see any".
>
> I did not say "what tie project?", I said "What _is_ the tie project?".
> That does not mean that I am of the opinion that nothing has been done,
> but that there is no specific objective spelled out that would be a
> suitable target for a limited-time project.

David, you miss the point and nitpick on wording.
I suggest that you start focusing on the actual problem instead, if
you want to improve the situation.

>> And i didn't only collect lists of ugly ties.  I already have a spec
>> draft, and a pretty good idea how to approach the issue.
>
> You have a "spec draft" and yet you claim that the project is
> well-defined in a manner where everybody should know what it entails,
> making the question "what _is_ the tie project?" an insolence?

Please use simpler English as i have trouble understanding what you mean.

>> Several times i have offered to share my research.  Usually it seemed
>> that noone cared to see it.
>>
>> So, I am really angry because of what you said and how you said it.
>
> So if what I had written was so foolish, how about you write a short
> GSoC proposal spelling out specifically what a student is supposed to
> achieve exactly in the three months assigned to the project?  There may
> be something like 2 to 4 hours left before the proposal time window
> closes.

What a sudent is supposed to achieve:
- design a new formatting strategy that will format correctly all
given tie samples (Test-driven development), based on my research
results
- design the code to be flexible so that one can supply different
settings to get differently formatted ties (according to his
preference, but within limits of what is acceptable typogaphically),
and easy to override a particular tie appearance
- write code that is maintainable, commented and tested (my examples
should be good regression tests)
- as a bonus, he may write support for enharmonic ties and ties
spanning clef changes.

Feel free to expand/adjust/finish this - for me it's too late, as i
have to go to my day job today and i must sleep at least 8 hours
before that or i'll be useless.

Janek