On Mon, 17 May 2010, Unidata User Support wrote:
Hi Gilbert,
Hello Tom,
The portion of code that was vunerable to bad products is in GRIB2
processing. In particular, products that were suppoed to be individual
GRIB2 messages but were not were sent to g2clib routines for metadata
extraction. The problem was that g2clib assumed that the GRIB2
messages are error-free -- they did no sanity checking on their
size/contents. Since the product's size is available before g2clib is
called, it was straightforward to pass the product size to g2clib and
modify the g2clib code to check the indicated product size with what
was read and not proceed with badly formed messages.
OK. So in other words, if the same thing happens with NIMAGE products, or
IDS|DDPLUS, it shouldn't crash the noaaport software because it's just
being "passed on" instead of going through g2clib. Is this what I am to
conclude?
re:
BTW, College of DuPage and I (NIU) have been running it for about 8 hours
now with no problems. I was wondering, however: I indirectly inject into
the IDD through Unidata2@SSEC. Since I don't inject directly, am I safe to
stay on this version, or should I revert back to 1.3.0 until Tuesday
night?
I checked the one machine I could find that is requesting data from
your NOAAPort ingester and saw that NEXRAD3 is not being requested.
Given this, there is no chance of the DPA products "leaking" into
NEXRAD3 before users are alerted to the change. So no, you should not
need to revert.
OK. Thanks again for the update!
*******************************************************************************
Gilbert Sebenste ********
(My opinions only!) ******
Staff Meteorologist, Northern Illinois University ****
E-mail: sebenste@xxxxxxxxxxxxxxxxxxxxx ***
web: http://weather.admin.niu.edu **
*******************************************************************************
