ExaBGP 'Homeland Security' release

This version is a bug fix only release.

Thank you to everyone who took the time to test and report issues. As a result the changelog is pretty long, curing mostly obscure bugs with one large RIB code change fixing three of the issues reported at once. It is nice when it happens unexpectedly :-)

The CHANGELOG lists :

• fix RIB cache handling
• fix flow generation (mandatory attributes are mandatory even for Flow Routes), thanks to: Quentin Loos for reporting the issue and helping fixing it
• fix an issue on SIGUSR no route withdrawal update was performed, reported by: Sascha Schumann
• Do not send routes on route refresh for unrequested families, reported by: Hideaki HAYASHI
• Handle gracefully withdrawing of absent routes with the API, reported by: Peter Bristow
• Correct malformed JSON message, reported by: Robert Barnes
• Validate the FLAG of the attribute received (following our own advise on IDR)

This version was released exactly 12 years after the creation of the department of Homeland security ... Promise it is just a coincidence.

ExaBGP 'internap' release

A bug was introduced in the JSON code of 3.2.14 and was fixed with this release.

ExaBGP 'google' release

This version of ExaBGP fixes a critical bug, all user of previous 3.2 release are asked to upgrade as soon as practical.

In this release :

• Fix: do not leak route between peers
• Fix: restore group level static group for all peers thereafter
• Fix: a bug in group-updates causing attributes to not be included in the update message
• Fix: a bug when update large than 4096 could be generated with group-updates
• Fix: an issue with JSON generation
• Fix: MD5 support had been removed by error from 3.2
• Improvement: allow to use quote with --decode
• Improvement: implement draft-ietf-idr-aigp-10

ExaBGP 'release early, release often' release

This version is a bug fix with only, mostly to not eat all the CPU of a computer when connecting aggressively.

Hopefully the pace will now slow down a lot. I really embraced "release early, release often", even too much to my own taste.

I am only aware of one issue for someone with some very high performances requirements, which I could not reproduce yet.

ExaBGP 'oops' release

Fix a critical bug in 3.2.11 (and perhaps previous 3.2 release) when routes would not be resent on reconnection to a lost peer.

All 3.2 users are advised to upgrade.

ExaBGP 'Pirate Yesterday' release

This release is mostly for bug fixes, one major exception being the implementation of draft-keyur-bgp-enhanced-route-refresh-00.

• Fix: bug in generating API string, missing space
• Fix: bug in generating API string, withdrawn reported as invalid announced
• Fix: bug in generating API string, withdrawn was missing Path-Information
• Fix: a bug with capability parsing introduced with operational
• Fix: be more aggressive on reconnection (3.2 was slower than 3.1)
• Improvement: the TCP server allows 'neighbor' commands
• Improvement: enhanced route refresh initial implementation

ExaBGP 'KEF' release

This release possible thanks to Keflavik Airport's free wifi, adds support for RFC 2918 - Route Refresh Capability for BGP-4 and fixes two issues.

ExaBGP 'operational' release

A bug fix release with some partial draft-ietf-idr-operational-message-00 support.

ExaBGP 'on time for the EPF' release

Include a small fix which was left in the stable branch and did not make it up to master ...

ExaBGP 'dead on the day' release

As no one reported any issues with the pre-releases.