From: Jacob <jacob01_at_gmx.net>
Date: Sun, 17 Oct 2004 23:40:05 +0200

Hello,

On Thu, Oct 14, 2004 at 01:25:54PM -0700, scott wrote:
>
> --- Neon John <johngd_at_bellsouth.net> wrote:
> >
> > With >8000 files on my 80 gig machine I've never felt the need for
> > anything more than the simple file browser. What am I missing?
>
> What you're missing is that _the majority of people_ aren't nearly so
> organized. My music folder contains ~12,000 files. It also contains ~1200
> folders. I very rarely have problems with finding songs, the exception being
> looking for a song that I know is on a soundtrack, but don't know which one,
> but even then I can usually narrow it down to less than 5 choices. Most people,
> > >3) How to synchronize/generate the database?
> >
> > I think a line should be drawn in the sand against any dependence on a
> > host computer. The archos should be able to do any database manipulation
> > without being tethered to a computer. Requiring a desktop-based
> > application would put Rockbox down in the same class with the neuros, ipod
> > and all the other crappy tethered players. I have no problem with there
> > being a PC application that duplicates what the archos does, only faster.
>
> I don't think this is necessary.
> The process would take forever on an archos.

I don't agree to this hypothesis. Instead I would like to recall the
dbase argument: Back in the good old days (TM) it was possible to index
this amount of data (say 12000 records) on slower CPUs with much less
RAM and much slower mass storage in reasonable time. To achieve this
they've probably put a lot of thoughts into their file formats.
Unfortunately I'm not a database developer, but I suspect at least the
following:

  - Having fixed-length records makes it much more efficient to do
    inplace updates or re-sorting of index files that dont fit
    completely into RAM.

  - Having multiple files allows it to add records (either by appending
    or by replacing) without inserting data into a file, which is a
    costly operation, especially if you do it at the beginning of a 20M
    database file (ok, this can be circumvented (1) by FAT-level hacks and
    stuffing bytes to cluster borders or (2) by having a huge file with
    a fs like structure and internal fragmentation).

  - Compared to the CPU/RAM limitiations, there is a lot of mass storage
    accessible, so we don't have to optimize the file format for size,
    but we can focus on fast searches and reasonably fast re-indexes or
    updates.

  - BTW, can't we use the full MPEG buffer at least for the
    re-indexing process?


> What separates the archos from the neuros and iPod is that they _require_ an
> ID3 database. With rockbox, if you don't want one, you don't have to have one.
> If you DO want one, you're probably willing to go to the trouble of building
> the DB on your computer (and as another post said, you need to connect to your
> PC to add music anyway).

Hmm, I do a lot of recordings without having access to a host meanwhile,
so an external application (even if it were coded in python or perl and stored
on the JBR) wouldn't be a help here.
>
> -scott
>

Having an efficient file format would make it possible for all to use
this feature, starting with an adhoc one would probably restrict it to
the host oriented ones for some time (until it gets replaced).

Just my 5c,

Jacob
_______________________________________________
http://cool.haxx.se/mailman/listinfo/rockbox
Received on 2004-10-17