[mythtv] MythMusic: MetaLibrarian RFC

[Colin Guthrie]

thor wrote:
>  In the abstract, it's a fine idea. We don't all use parochial code to do host 
> name resolution; we all ask DNS to resolve host names for us.  Why not a 
> similar service for audio metadata, even if only at the local level.

Thanks, glad you agree on that level :)

>  But describing an idealized service is one thing. Building it is quite 
> another. Getting a non-trivial percentage of applications to even be aware of 
> its existence is a still much larger hurdle. 

Yes I realise this. I belive I can code this system relativly easily. 
The principles are fairly standard and there are big chunks of programs 
I/others have previously written that could deal with large portions of 
this project too.

The changes to Myth would be easy enough to code, albeit take a little time.

Other programs etc. may be harder to convince, and it may take a fair 
bit of time to gain acceptance, but everythings got to start somewhere.


>  If you want to get anywhere close to a path that might lead to wide 
> acceptance of your idea, you will need to make it work with current 
> (unmodified) software. That means, to the extent possible, pushing metadata 
> back into the mp3's themselves (ie. writing tags that other software can 
> read).

Oh yes. I totally agree. Simon also seemed to hint at this, but I am not 
suggesting that it does not write tags to files. If running as a daemon, 
apps could submit metadata changes to the MetaLibrarian and it could at 
a later date (or immediately) write these changes to the file itself. 
I'll rewrite that part of the description to make it clearer that file 
based tags will *always* be maintained.


>  So ... uhm ... for mythmusic, where exactly are we on 2.3 versus 2.4 tags? As 
> I understand things, we're now writing tags that almost nothing else can 
> understand. Is that accurate? If so, can we improve on that  :-)   ?

Well, my general thought was that if the metadata writing was off loaded 
to another program/library then it could use a better tag writing 
library to write to the files and thus get round the whole problem! It 
doesn't really solve it directly tho' and would still add other 
dependancies (although I reckon such dependancies could be wrapped up in 
to a libmyth component, to keep it "in house" kinda).

I decided last night to write a simple ID3v2 tag writer for Myth and use 
that for writing tags. The spec is uber simple to follow so it will only 
take a few days of coding to get it right and libid3tag can P155 right 
off. It's caused me enough hassle. (OK, it can stick around for reading 
tags - that at least seems OK ;)

> btw: DAAP does have rather extensive metadata-description capabilities, but 
> in-and-of itself it is just a markup and streaming protocol. It says nothing 
> about how that data should be persisted.

Interesting, thanks for the summary. So a DAAP daemon could still use 
this as a backend for persistant storage of the data etc.

Cheers for the feedback.

Col.
-- 

+------------------------+
|     Colin Guthrie      |
+------------------------+
|  myth at colin.guthr.ie   |
| http://colin.guthr.ie/ |
+------------------------+