rietveld should hide datastore timeouts when possible

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Use rietveld a lot.
2. Notice a few 500 errors a day due to datastore timeouts.

What is the expected output? What do you see instead?

This article
http://code.google.com/appengine/articles/handling_datastore_errors.html
says that datastore timeouts are a fact of life on App Engine.
Rietveld should expect and handle them gracefully, both to make
the Rietveld experience better and because it is often looked to
as a model App Engine app.

It would be much nicer for the request to take, say, ten more seconds
than to put up a 500 error page.  The page linked above suggests
that retrying shouldn't help, but most of the errors I see go away when
I retry manually, which suggests an automatic retry would be worthwhile.

I also get 500 errors when making comments.  These are nicely handled
by putting up a little error box where the comment should be, but then
I've lost the comment text and have to type it again.  It would be
even nicer if the "writing a comment" box stayed up and the javascript
retried in a second or two.

Right now hardly a day goes by that I don't see a transient Timeout
error.  Rietveld is a fantastic tool, and these timeouts detract
from the experience considerably.

Original issue reported on code.google.com by rsc@golang.org on 23 Dec 2009 at 4:00

[GoogleCodeExporter]

Original comment by albrecht.andi on 23 Dec 2009 at 8:13

• Changed state: Accepted

[GoogleCodeExporter]

In r57 the code to automatically retry was disabled but there is no comment on 
why. Can 
Guido comment on the rationale behind this as he did the change? I haven't 
found 
anything related in the ML archives.

Original comment by maruel@chromium.org on 3 Feb 2010 at 8:22

[GoogleCodeExporter]

Original comment by albrecht.andi on 6 Apr 2012 at 6:31

• Added labels: Component-Server