Re: GNOME_VFS_OPEN_RANDOM

[Alexander Larsson <alexl redhat com>]

On Mon, 2006-11-20 at 15:42 +0100, ensonic wrote:
> hi,
> 
> there is http://bugzilla.gnome.org/show_bug.cgi?id=354590. Its about a
> property to tell gstreamer gnomevfs source to use GNOME_VFS_OPEN_RANDOM or
> not. Now people are quite against adding that. None other know filesystem
> needs a flag like this.
> 
> When opening an http uri, after the first get request gnomevfs should know
> if seeking is possible. If not seek calls can be made to fail. If there is
> a seek without a previous get, you will probably need to get 1 byte and
> keep that, so if seek is not possible and the app next reads 10, bytes
> you'll read only 9 and put the one already read in there first.
> 
> if that totaly makes no sense, any ideas about the bug?

In some cases its possible to do a better job if you know that there
will never be any seeks. You might for instance have to do some ugly
hacks to implement it that hurts the streaming case. For instance, seeks
and writes to an open handle is complicated in many backends.

GNOME_VFS_OPEN_RANDOM also unfortunately have some ugly API entaglement.
It also defines if opens for write truncate or not, and its also used as
a hint for streaming optimizations.

Anyway, what you propose, an open for random access not failing even
though random access isn't supported would be an API/ABI break, so that
won't happen.

I don't see why you can't just try again without random if it fails
though. How is that different from failing in the seek later.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
 Alexander Larsson                                            Red Hat, Inc 
                   alexl redhat com    alla lysator liu se 
He's a suicidal drug-addicted gentleman spy from a doomed world. She's a 
psychotic thirtysomething mermaid who can talk to animals. They fight crime!