Hi Yannick,
I've received a few complaints on this issue also. The trouble
appears to be associated with very long connections to retrieve a
single file. Particularly, if the amount of time for the connection
exceeds 3600 seconds for a single file.
I do not have the ability to detect if the requested file has been
corrupted en-route or is just being accessed incorrectly. So, we
have a problem. If we provided a checksum method then users could
see if the file they are getting matches what we sent. Such a system
should be implemented in a generic way for the archive.
I'll try and discover how frequently these sort of transfers are
failing.
JJ
On 20-Aug-07, at 10:44 PM, Yannick Mellier wrote:
>
>
> Dear JJ
>
> It seems several users complained, or were lost, with the
> compressed data. Below, an email from Jacqueline Bergeron that you
> received by aug. 10. Jean-Francois Sygnet got some problems as
> well, as far as I understand? (we did not received complains but
> we got feed back from one colleague). I am not sure it is our
> fault, I think Sygnet retrieves data from a script written for
> T0003 and did not check T0004 used another compression algorithm.
> Do you have an way to send an automated message to users in case of
> failure during the upload that warm people they check they ask for /
> gz files?
>
> Several still believe there are RICE compressed, I guess. The
> Terapix pages warmed users about the gz compression for T0004, but
> I am not sure users really read documents:
>
> http://terapix.iap.fr/article.php?id_article=715
>
> (first item, last line in bold)
>
> May be we should make it even clearer that T0004 images are gzip-
> ed compressed?
>
> Cheers
>
> Yannick
>
>
> email from Bergeron:
> -------------------------
> Dear Colleague,
>
> After downloading several fits images this week without problem,
> one file seems to have an invalid format :
> gunzip CFHTLS_W_z_022150-041200_T0004.fits.gz
> gunzip: CFHTLS_W_z_022150-041200_T0004.fits.gz: invalid compressed
> data--format violated
>
> Thanks for your help to get the correct gz fits image for this
> field and colour
> (the i and u images are fine).
>
> Jacqueline Bergeron
>
Received on Tue Aug 21 2007 - 06:00:54 HST