On Thu, Aug 08, 2002 at 03:13:37PM +0200, G?ran Thyni wrote:
>
> On Wed, Aug 07, 2002 at 10:10:29PM +0200, Oskar Sandberg wrote:
> > On Wed, Aug 07, 2002 at 06:56:43AM -0700, G?ran Thyni wrote:
> > > It probably does NOT work at all!!
> >
> > I'm afraid the final statement is probably the truth here.
>
> Great,
> I checked-in early to get comments so I am happy you looked at it.
>
> > In order for the Buffer implementation to do it's job, it needs to
> > support "sliding locks". That is to say: if somebody starts reading a
> > buffer while is being written to, you need to make sure that the reader
> > can read as much as possible, but that it stays behind the writer by
> > wait()ing when it caches up.
>
> One approach would be the simplistic way to get
> a atomic write:
> - Write to tmp-file then "File.rename()"
>
> at least on unix an open file is refer to by inodes so
> as long as the file is already open you can read until
> you close it, changes in the directory has no effect.
>
> Still it may be best (but not necessary) to block multiple
> writer.
>
> Problem with this approach?
> What about non-unix systems, anyone?
>
> > It also needs to do vice versa: if somebody
> > starts overwriting a buffer while it is being read from, you need to make
> > sure that the writer can write as much as possible but then wait()s when
> > it caches up with the reader.
>
> Do we ever rewrite keys? Not CHK keys?
> I will have to read up on the key type and their function.
Never ever ever at the moment.
>
> regards,
> --
> G?ran
>
>
>
>
>
> _______________________________________________
> devl mailing list
> devl at freenetproject.org
> http://hawk.freenetproject.org/cgi-bin/mailman/listinfo/devl
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL:
<https://emu.freenetproject.org/pipermail/devl/attachments/20020808/20c9d903/attachment.pgp>
