Rani,
thanks, but this is not the problem I am trying to counter. I have
performed downloads in the past where there was actual *corruption* of
the file transferred. Where the site also provides a telnet login I have
solved this by logging in, splitting the file into 100k lumps, taking
checksums and then ftp'ing in to get all the lumps. I can see if any of
the lumps were corrupted by checking the checksums. I then only have to
re-get those lumps that were corrupted.
I acknowledge that 100k lumps are prolly a bit small to make generally
available, but 1M lumps should be ok.
Cheers,
Wilf
> ----------
> From: Rani Pinchuk[SMTP:[EMAIL PROTECTED]]
> Sent: 01 June 1998 10:17
> To: Coles, Wilf
> Cc: Java-Linux
> Subject: Re: Size of releases - compression/deltas
>
> Instead of doing a complete download if something went wrong, I think
> it is
> better to use the ftp restart command (from the ftp man):
> restart marker
> Restart the immediately following get or put at the
> indicated
> marker. On UNIX systems, marker is usually a byte
> offset into
> the file.
> which means that if you have a the first X bytes of a file, you should
>
> do 'restart X' before you try to get the file again.
>
> Rani.
>
>
> On Mon, Jun 01, 1998 at 08:09:26AM +0100, Coles, Wilf wrote:
> > My plea would be for all of that :-) and also to split the download
> into
> > a number of checksummed smaller chunks. At least then any corruption
> > doesn't mean a complete download again.
> >
> > Cheers,
> > Wilf
> > ...and thanks for all the good work!
> >
> > > ----------
> > > From: Artur Biesiadowski[SMTP:[EMAIL PROTECTED]]
> > > Sent: 31 May 1998 18:37
> > > To: Java-Linux
> > > Subject: Size of releases - compression/deltas
> > >
> > > I'm probably not only person with slow, not reliable inet
> connection.
> > > Any way to reduce download size of jdks would be very nice. I
> think
> > > about two things:
> > > 1) Use bzip2 instead of gzip. It is normally about 20% smaller (it
> > > counts whith files > 10MB). Of course gzip is more popular, but
> > > providing either link or bzip binaries wouldn't be a great
> problem. Or
> > > if space allows distribute both .gz and .bz2 versions.
> > >
> > > 2) Use xdelta (from ftp://scam.xcf.berkeley.edu/pub/jmacd)
> > > It generates binary diffs between archive files - I suppose that
> it
> > > could be very efficient for small version differences - ie from
> > > 1.1.5-v6
> > > to 1.1.5-v7. Could somebody with access to all these jdks make
> tests
> > > and
> > > report it ? Maybe even cross versions could be done (1.1.5 to
> 1.1.6).
> > >
> > > Artur
> > >
> > > P.S.
> > > I would be personally interested in xdeltas of jdk - I do have
> > > linux-jdk.1.1.3-v2.tar.gz size 9093801
> > > Can somebody make xdeltas to one of new releases (newest 1.1.5 or
> > > 1.1.6)
> > > and report their sizes/mail them to me if smaller than 2-3 MB ?
> > >
> >
> >
>
> --
> ----------------------------------------------------------------------
> --------
> Rani Pinchuk <[EMAIL PROTECTED]> Tel: +32-3-326-79-97
> http://www.kinetica.com/home/pinchuk
> ----------------------------------------------------------------------
> --------
>
