Re: learning about daemon-free metadata

[Seth Nickell <snickell stanford edu>]

Hi Darin,

I don't *have* a daemon-free metadata proposal. What I'm trying to
understand is why we consider it unacceptable (or at least a bad idea)
for metadata to be lost by resource synchronization issues, but losing
"normal data" seems acceptable.

The only thing I can think of is that while we understand metadata for a
single file as an object (because we know the format, etc), as far as
the file layer is concerned files on the disk are black boxes that get
changed in mysterious ways. This seems to be, more than anything, a
result of historical development and not a limitation we should be
inherently assuming.

ATM I'm planning to implement metadata using a daemon.

-Seth

On Sat, 2002-06-29 at 10:40, Darin Adler wrote:
> Hi Seth, I have two questions for you about your daemon-free metadata 
> proposal.
> 
> 1) With your proposal, if one application says:
> 
> 	setMetadata("/this/is/my/filename", "icon", "dog");
> 
> and another application says:
> 
>           setMetadata("/this/is/my/filename", "preferred application", 
> "gimp");
> 
> at the same time, are both of these changes guaranteed to be made? Or 
> might one be lost?
> 
> 2) With your proposal, if one application says:
> 
> 	setMetadata("/this/is/my/filename", "tiny-thumbnail-jpeg", 22K of 
> image data in JPEG format);
> 
> and another application says:
> 
>           setMetadata("/this/is/my/filename", "tiny-thumbnail-jpeg", 14K 
> of image data in JPEG format);
> 
> at the same time, are you guaranteed to get either one set of icon data 
> or the other? Or could you end up with something that's half of one and 
> half of the other and not valid JPEG?
> 
> The answers to these two questions may help me understand your proposal.
> 
>      -- Darin
>