Re: Update to RIS raw data format for 4-byte ASNs
Hi Erik, I didn't follow any of the discussions on the new MRT format, as I (naively) thought it would just be minor update for the longer AS numbers. But after reading your recent mails and having started work on importing RIS raw in a SG project, I see there's a fundamental change (at least for RIB dumps): in the old format, a call to bgpdump_read_next would return information on one and only one entry of the routing table. In the new format, a call to bgpdump_read_next returns a whole block of entries; for the latest rrc03 bview, anything between 1 and 35 at a time. To me it looks strange, seems to make programmer's life needlessly more complicated. Now one not only has to loop over all bgpdump_read_next records, but also loop over all entries in each "record". Programs which worked with old raw data will need major revision; and those which need to work with both old and new format will look ugly, a different paradigm for processing old and new data. Do you have any idea of the reasons behind this big change? Cheers, -- Rene
Dear Colleagues,
This is an update concerning 4-byte Autonomous System Numbers (ASNs).
An IETF draft has been released describing a new version of the MRT format, used for RIS raw data. We have published a new version of libbgpdump to support this, which can be downloaded from: http://www.ris.ripe.net/source/
There is at this stage no set date on which 4-byte ASNs will appear in the RIS. The current raw data does not use the new format.
Details of the changes to RIS for 4-byte ASNs, including more information on the new raw data format, can be found at: http://www.ripe.net/projects/ris/docs/asn.html
If you have any questions, please send an e-mail to <ris@ripe.net>.
Regards,
Erik Romijn Information Services Department RIPE NCC
On Wed, Mar 28, 2007 at 11:10:43PM +0200, Rene Wilhelm wrote:
But after reading your recent mails and having started work on importing RIS raw in a SG project, I see there's a fundamental change (at least for RIB dumps):
Yes, that's true. Updates are indeed pretty much the same.
To me it looks strange, seems to make programmer's life needlessly more complicated. Now one not only has to loop over all bgpdump_read_next records, but also loop over all entries in each "record".Programs which worked with old raw data will need major revision;
It's a bit more work, yes. It would be additional work anyways. And IMHO, with the provided examples, it's really not hard at all.
Do you have any idea of the reasons behind this big change?
Yes. 38M Mar 28 08:00 /export/data1/rawdata/2007.03/bview.20070328.0800.gz 21M Mar 28 16:00 /export/data1/rawdata/2007.03/bview.20070328.1600.gz (rrc03) cheers, -- Erik Romijn RIPE NCC jr. software engineer http://www.ripe.net/ Information Services dept.
Do you have any idea of the reasons behind this big change?
Yes.
38M Mar 28 08:00 /export/data1/rawdata/2007.03/bview.20070328.0800.gz 21M Mar 28 16:00 /export/data1/rawdata/2007.03/bview.20070328.1600.gz (rrc03)
ah, that's a good reason. Must be a lot of overhead in the old format! -- Rene
Rene Wilhelm wrote:
Do you have any idea of the reasons behind this big change? Yes.
38M Mar 28 08:00 /export/data1/rawdata/2007.03/bview.20070328.0800.gz 21M Mar 28 16:00 /export/data1/rawdata/2007.03/bview.20070328.1600.gz (rrc03)
ah, that's a good reason. Must be a lot of overhead in the old format!
Hi Erik, as a user of libbgpdump, I'm curious: does this change break existing users of the MRT format, even if they don't care about 32-bit ASNs? I'm sure you don't want people who download the RIS raw data and try to parse it with route_btoa or old versions of libbgpdump to stop being able to use the raw data files, right? If this stuff did break, were the users warned with an appropriate email to the ris-users mailing list? I didn't see one, but admittedly I only briefly scanned the archives... Cheers, Lorenzo -- --------------------------------------------------------- Lorenzo Colitti Ph.D Computer Networks research group Roma Tre University colitti@dia.uniroma3.it +39-0655173215 ---------------------------------------------------------
[ I can't cc to sg@ripe.net, unless someone approves it in mailman ] On Fri, Mar 30, 2007 at 02:10:24PM -0700, Lorenzo Colitti wrote:
as a user of libbgpdump, I'm curious: does this change break existing users of the MRT format, even if they don't care about 32-bit ASNs?
Yes for RIB dumps. For update dumps, you just won't see any updates from a 4-byte AS peer. For RIB dumps, the entire format changed, so you won't see anything with older versions.
sure you don't want people who download the RIS raw data and try to parse it with route_btoa or old versions of libbgpdump to stop being able to use the raw data files, right?
I prefer not to, but IMHO, it was the best option here. The other option could be to dump only route entries with a 4-byte AS in the new format. But that would lead to a very very nasty quagga patch and people might be missing routes without knowing.
If this stuff did break, were the users warned with an appropriate email to the ris-users mailing list? I didn't see one, but admittedly I only briefly scanned the archives...
Yes. One on Jan 11, one on Mar 15 with details of the new format, one on Mar 22 (also sent to routing-wg) to say that changes will be applied on Mar 27. cheers, -- Erik Romijn RIPE NCC jr. software engineer http://www.ripe.net/ Information Services dept.
participants (3)
-
Erik Romijn -
Lorenzo Colitti -
Rene Wilhelm