So, it looks like we've generally included two change listings in our releases: ChangeLog and NEWS. It's pretty obvious what NEWS is supposed to be: a human edited list of the most important changes in the release. There is some question, though, about what a ChangeLog should be. Right now, it is useless, as nobody updates it. Who wants to write two commit messages rather than one, afterall? If we actually want a change log separate from an "important changes" NEWS file, I think it'd be fine to simply autogenerate it from our SVN log messages, because if you want a more concise explanation of what changed, that's what NEWS is for. The advantage of doing this is that people who don't follow SVN get to see all the things we changed, the disadvantage being that it's 6500 lines long and contains some irrelevant parts. E.g. merging trial, reverting merge, merging trial, reverting merge, repeat like 200 more times. Also, things lilke "Use ThreadedResolver instead of BlockingResolver by default" in one commit, and "Tra la la la la, like extra flowers" in the next when you realize you screwed up the first and need to do a 4-character fix. :) On the other hand, perhaps all we really want is a single human-edited changes file, describing in a bit more detail than the 1.3 NEWS file what's changed, but more concise than a full changelog. What do people think? James
On Thu, 10 Mar 2005 16:13:26 -0500, James Y Knight <foom@fuhm.net> wrote:
On the other hand, perhaps all we really want is a single human-edited changes file, describing in a bit more detail than the 1.3 NEWS file what's changed, but more concise than a full changelog. What do people think?
+1, sounds good to me. -- Twisted | Christopher Armstrong: International Man of Twistery Radix | -- http://radix.twistedmatrix.com | Release Manager, Twisted Project \\\V/// | -- http://twistedmatrix.com |o O| | Founding Member, Hobart Hacking Society w----v----w-+ -- http://hackingsociety.org/chapters/hash
Christopher Armstrong wrote:
On Thu, 10 Mar 2005 16:13:26 -0500, James Y Knight <foom@fuhm.net> wrote:
On the other hand, perhaps all we really want is a single human-edited changes file, describing in a bit more detail than the 1.3 NEWS file what's changed, but more concise than a full changelog. What do people think?
+1, sounds good to me.
I agree. If people want more detail than is in NEWS, they should probably just subscribe to the commits list anyway (as I did), or they can access the svn logs themselves as desired. Steve
James Y Knight wrote:
On the other hand, perhaps all we really want is a single human-edited changes file, describing in a bit more detail than the 1.3 NEWS file what's changed, but more concise than a full changelog. What do people think?
What I do with ldaptor is take the SVN commit messages, clean them up, categorize and prioritize. People who want just high-level can read the first section only, people who are only admins/users can read the first two sections, and developers can pick the parts they are interested in. I also take care to trim out anything that isn't relevant -- e.g. things that are a continuation of work committed after 0.0.30 was released. On the other hand, I don't bother providing a NEWS file, atleast for now. Looks like this: ldaptor (0.0.31) unstable; urgency=low * Conceptual changes: - Service-specific passwords: added a mechanism that allows having "subpasswords" that are only used with particular services; e.g. let the userPassword attribute be used only for SSH logins, and use additional "mail" and "web" password, stored in an attribute called "servicePassword". The value is two whitespace-separated fields: service name and userPassword-style password digest. See ServiceBindingProxy for more. - LDIFTree allows LDAP server data storage an a one entry per file format, with nice unixy semantics. * Applications changes: - WebUI: - Add pretty label to search form submit button. - Service-specific password editing support. Closes: ldaptor#17. - Server: - New LDIFTree storage backend, storing the LDAP as a filesystem directory tree, with one entry per file in LDIF format. - Add LDAP proxy support. - Add ServiceBindingProxy, an LDAP proxy that handles non-anonymous bind requests specially. Closes: ldaptor#15. * API changes: - Get rid of MutableString: - Add pureber.berDecodeObject, remove pureber.ber2object. - Add pureber.berDecodeLength, remove pureber.berlen2int. - Switch LDAPClient and LDAPServer buffers to be plain strings. - Remove MutableString. - Change the BER decoding mechanism: - Add classmethod fromBER instead of using __init__ in two roles. - Remove old-style decode support. - Add pureber.berDecodeMultiple, mainly for use by subclasses of BERSequence. - Large lengths appeared to be negative, switched berlen2int to using longs. - Split LDAPServer to BaseLDAPServer, so one has a choice of inheriting something without any handle_* methods. - Make ldapsyntax.LDAPEntryWithClient.move() remove the old RDN. - Make ldapsyntax.LDAPEntryWithClient.move() change the entry's idea of its own DN. * Refactorings to adapt to Ldaptor APIs: - Use connectAnonymously in parallelsearch example. * Refactorings to adapt to Twisted/Nevow APIs: - Nevow ISerializable is deprecated. - ctx.{allPatterns,patternGenerator} are now called ctx.tag.allPatterns etc. - all IRequest APIs take a context. - switch from getDynamicChild to childFactory. - addSlash where appropriate. - return URLs instead of explicit redirects. - switch from renderHTTP to child_. - switch from locateChild to childFactory. - Twisted >1.3 moves SRVConnector to twisted.names.srvconnect. * Bugfixes: - Fix BER length encoding for lengths greater than 127, add unit tests. - Make many __repr__'s work even when __init__ failed. - Make sure every LDAP wire class takes a tag argument. - Make LDAPDelRequest also understand keyword argument value, as that is what LDAPString uses. - Fix and test handling of unknown match types. - Fix and unit test saltless non-client setPassword. - Fix DistinguishedName order comparison. - Make sure configuration does not interfere with the lmPasswords in the unit tests. - Make nevow data access to LDAP entry attributes work. - Add items() to LDAPEntry IContainer adapter's attributes child; webui search needs it. - Webui edit can now edit attributes that are in the RDN. Closes: ldaptor#5. - Improve handling of missing baseDN in configuration. * Cleanups: - Remove dead code. - Remove the hack-and-slashed RFC file that served as a TODO note. - Override IOPump to pass around more than one byte at a time, saves >10 seconds in unit tests and protocol decoding of partial messages is unit tested elsewhere anyway. - Improve ldaptor.testutil error messages, accept any true/false value and not just 1 and 0. - Clean up LDAPServer.handleUnknown. - Prettify LDAPMessage.__repr__ - Prettify LDAPResult.__repr__ - If ldapsyntax.LDAPEntryWithClient.search gets a bad response from the LDAP server, raise a real exception and not an assert. - When converting delta Modifications to LDAP, convert them to basic data types for easier debugging and unit testing. - Be consistent and use "ctx" everywhere in webui change_password. - Add nevow renderer that produces zebra tables. - Make ldapsyntax.LDAPEntryWithClient.move() accept strings etc. as newDN, add unit test. - Make epydoc happy. - Simplify ldap-intro document generation. * Packaging: - Depend on Nevow v0.3.0 or later. - Depend on Twisted v1.3.0 or later. - Switch to CDBS. - Add an ldaptor-common package to contain conffiles. - Run unit tests before building debs. - Include the unit tests in the distributed files. * Administrivia: - Add TODO notes. - Update ignore file lists to match currently generated files. - Export dia files to format png-libart, not just png. This makes the export reliable. Hint from Debian bug#275959. -- Tommi Virtanen <tv@debian.org> Fri, 10 Dec 2004 17:39:36 +0200
participants (4)
-
Christopher Armstrong -
James Y Knight -
Stephen Waterbury -
Tommi Virtanen