[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[leafnode-list] call for help: Managing leafnode's future - priorized TODO



Hi,

I've sorted an priorized the leafnode TODO list. I believe it's not
current everywhere, particularly, I did not review the 3.*, 4.* and 5.*
items.

Since I've experienced my spare time will be much lower in the near
future, I'll have to priorize my work and do more planning than just
random hacking of the past.

I'm now asking for your help, I plan to keep this TODO list available on
the web, tracking both the official version and my version in one
file. (I'd tag things like "fixed in 2.0b8_ma1" or something), and to
give us a clean start, *please* do review this list and tell me what
things are fixed and what are still open. 

Is there anybody who has experience with both Aegis and CVS or other
revision control and configuration control systems, freely available
ones? Pointers to comparisons on the web appreciated.


What bug tracking systems are good and which allow to priorize work,
possibly with incident tracking/tickets? Of course, it must be freely
available and should not have a track of security incidents itself (like
Mailman or stuff). Don't tell me anything about jitterbug, that one's
intransparent, inconvenient crap. Waste of time, a mail alias does a
better job than this one.



If there are any topics that are already addresses and can be dropped
off the list, if there are things missing, please do report these. Below
the line is a snapshot copy of my current leafnode TODO. It has
actions/suggestions on how to fix certain issues, feel free to
comment. If there's a language you don't know, it's most likely
German. ;) 

Please reply to the list if at all possible, if you think your English
is too worse, it probably isn't, but you can send things to me in German
or French as well although my French is not too good, it should be
sufficient to read. Please state if I may take your comment to the
public (list).

Expecting to be flooded away,

Matthias

----------------------------------------------------------------------

This is a PRIORITIZED LIST. It has two columns, first the importance,
second the urgence. (Matthias)



importance: (left column)
1 = most important, 5 = forget it
urgence: (right column)
1 = do right now, 5 = postpone until bored

allowed ranges:
  "security relevant": 1.1 ... 2.2
  "works correctly, but suboptimal": 3.3 ... 5.5
  "wish list": 4.4 ... 5.5



1.1 nntpd: POST may lose article if forked child fails. Must propagate
    error back!
    ACTION: Plan.

    Idea #1: discard the entire storearticle and lockfile and fixxover
    cruft, implement two queues (local and remote), should be inherently
    locking free (compare Maildir(5))

    Idea #2: make storage API to get around all that broken pseudo and
    local handling
    
1.1 groupinfo not updated after timeout + subsequent SIGINT
    ACTION: check if report still current

1.1 wenn connection nicht erfolgreich ist, haengt der rechner (?)
    ACTION: check if report still current

1.2 xoverutil: check whether tabs may be incorporated into .overview
    file
    ACTION: Discard TAB and replace by SPC if it's field separator.
    No games, no special cases.

1.2 don't supersede() articles unconditionally
    ACTION: figure criteria for allowing/denying supersede

1.3 authinfo handling in nntpd.c proper? (just user gives ACCEPT)
    ACTION: replace by proper state machine 

    Idea #1: Is there a good tool for generating code from state
    machines automatically?

2.2 articles are possibly lost when incorrectly linked (i.e. if
    /var/spool/news/message.id/xxx/<bla@fasel> is already present but
    /var/spool/news/news/group/name/nnnnnn is not).
    ACTION: check if report still current

2.3 "Das ist eine Macke von Leafnode. Wird von einem Server die
    komplette Gruppenliste geholt, so werden neu hinzugekommene Gruppen
    bei NEWGROUPS nicht aufgeführt. Der Reader findet sie daher nicht
    ohne weiteres." (Stefan Wiens, de.comm.software.newsserver
    2000-12-07, Message-ID: <wkr93lxerw.fsf@xxxxxxxxxxxxxxxxxxxxxx>)
    Transl: if a complete group list is fetched from a server, newly
    appeared groups are not offered in NEWGROUPS. 
    ACTION: fix. 
    Idea #1: Possibly emit list from mergegroups
    Idea #2: diff newsgroups list before and after group list fetch

2.3 if XHDR and XOVER fail, fall back to HEAD!!!
    ACTION: implement HEAD

2.3 flag servers as "don't post here" (Robin S. Socha, 2000-12-17,
    Message-ID: <3A3D1CB3.BeroList-2.5.9@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>)
    ACTION: implement server-specific configuration option

3.3 maxage doesn't work - still true?

3.3 fetchnews -N does not work
  Jürgen Salk, Message-ID:
  <3A3E8E47.BeroList-2.5.9@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>

3.3 run everything through various static and dynamic checkers
    (efence, dmalloc, lclint)

3.3 get rid of _all_ sscanf (Matthias Andree)

3.3 how to make newsgroup names 8-bit compliant? E.g., how to produce
  a newsgroups line of the form
	Newsgroups: norge.østfold
  Some readers encode this as ISO-8859-1 but this apparently is not
  understood upstream. It is also against the RFCs. son-of-rfc1036
  says that all data must be transmitted with ASCII < 127.
  draft-ietf-nntpext-base-07.txt says that the character set for all
  NNTP commands is UTF-8 (unicode!). Another possible draft is
  draft-ietf-usefor-article-01.txt (chapter 5.5.) which is not very
  helpful either.

3.3 fetchnews starts with an initialfetch (!)

3.4 postarticles is a horrible mess. It deletes the articles IMMEDIATELY
    if it's available on a server.
   
4.2 get rid of all those if(verbose > anything)

4.3 move old articles from out.going to failed.postings if they are,
    say, 7 days, old? (Report by Stefan Wiens, Message-ID:
    <3A1EDBFF.BeroList-2.5.9@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>)

4.4 can fetchnews cope with servers that reject XOVER? (Matthias)

4.4 signal needs to be worked around on Solaris 7 (Matthias Andree)

4.4 server dependent out.going, possibly a repost command?

4.4 interrupt fetch every 100 postings and then reconnect, to beat the
    INN configuration

4.4 nntpd Control-Nachrichten und Artikel in moderierte Newsgroups
    werden auch angezeigt. (Letzteres ist u.U. gar nicht so tragisch,
    denn es verhindert, dass Newbies ihre "moderierten" Postings
    wiederholt abschicken.)

5.5 filterutil.c: decode months with bsearch

5.5 can't find /usr/local-installed PCRE on FreeBSD 4.0 CFLAGS/LDFLAGS
    required

5.5 determine interesting groups from .newsrc?

5.5 delaybody only for certain groups?



-- 
leafnode-list@xxxxxxxxxxxxxxxxxxxxxxxxxxxx -- mailing list for leafnode
To unsubscribe, send mail with "unsubscribe" in the subject to the list