POS

Paul Prescod papresco@calum.csclub.uwaterloo.ca
Mon, 12 May 1997 13:46:19 -0400


Chris Bitmead uid(x22068) wrote:
> 
> >So my contention is that in the end most objects that are persistent
> >will be treated specially and "named" explicitly. The effort of giving
> 
> ls ~/.netscape/cache
> 00        05        0A        0F        14        19        1E
> 01        06        0B        10        15        1A        1F
> 02        07        0C        11        16        1B        index.db
> 03        08        0D        12        17        1C
> 04        09        0E        13        18        1D
> 
> I'm afraid this makes little sense to me I'm afraid.

What does "most" mean to you? How many *non-temporary* objects on your
hard disk use automatically generated names? Do a "find" on everything
except /tmp and you'll find that most things have human readable names.
 
> It is my contention that most persistent objects won't be treated
> specially. Only the high level objects that actually mean something to
> the casual browser will have a name in the traditional sense.

Right, that's the same as in traditional file systems. Every byte or
line in a .GIF is not named, only the gif file itself is named. But this
is a totally different issue: one of granularity, not naming.

The point is that in a non-transparent system anything that is saved to
the disk *must* have a name or be embedded in something else that has a
name. Even lazy developers do not have any choice about this. Since they
must give it a name anyone with half a brain will give it a *meaningful*
name because files named "01" and "1D" are logical targets for deletion
(which is why it is not a problem to name files in a cache this way). I
think that this is a good thing that makes system management workable.

 Paul Prescod