-*- mode: outline -*-
$Id: todo.txt,v 1.113 2003/09/08 00:56:41 mca1001 Exp $
This file contains the witterings of a loon. You don't have to
understand it if you don't want to. 8-)
The rest of my site is at
http://www.t8o.org/
http://toy.ddts.net/
and http://www.t8o.org/~mca1001/
It turns out this file fits the concept of a "programmer's notebook"
http://c2.com/cgi/wiki?ProgrammersNotebook
The guy is swamped, says "... huge backlog awaiting processing. If you want to help, you could become a Guest Editor." http://foldoc.doc.ic.ac.uk/foldoc/guest-edit.html Ooh, funky: http://foldoc.doc.ic.ac.uk/foldoc/source.html
1. To install the LILO bootloader
"I lilo'd [liloed] and rebooted"
2. To forget to reinstall the LILO bootloader after changing kernel
"I lilo'd the <a href>colo server</a> and had to get Hugh to
drive over and fix it"
Perhaps the second should be "lilofsck" instead? I've never used or
heard that word though, and Google returns no hits (2002-06-18).
(Oh a pointer to the howto might be handy)
3. To lounge around on a planar inflatable in or near a swimming
pool.
I lilo He lilo(e)s We lilo Or, if you're particularly embarrassed about such an accident which happened while you were sunbathing with your laptop, on the inflatable in your swimming pool... I had to lie low after liloing the colo from my lilo. 8-)
http://lists.debian.org/debian-user/1996/debian-user-199609/msg00159.html
I'm serving the stupid photo album because I can Problem is it kills my Starcraft game when I upload bulk stuff Having three altavista bots on the case makes a mess I've installed 'shaper' and I need to build the modules menuconfig is hanging, probably because I've broken libncurses or something. So, I'm off to mow the lawn. libapache-mod-limitipconn mod_bandwidth
~/public_html should be separate from my working copy, otherwise it'll break if I get a conflict from something I committed from another copy. Problem is, I'm frequently likely to hack on that 'cos it's easy to reach and 'live'.
- read about authbind - read about userv - install! Sending mail to message-id@domain is really pointless, especially when there are lots of them. I need a regexp that describes "likely to be a message-id".
I'm starting to hyperlink parts of my vapourware queue, as if they were permanent fixtures! They're not. This will cause trouble, especially since they're numbered.
Should allow linkcrawl to pull incoming referers and checking my content's bookmark names from there! - that thing should have a more descriptive mood than "rude mode" - should hyperlink the owner
When I lose the cable signal for a long time, the ddtcd thinks it's still online .. so it doesn't bother to tell anyone that it's back. (Later: pah, cron is a viscous solution. Too many running tasks when the link goes down.)
Someone can't mail me because his MTA barfs on my SSL config. > <<< 500 starttls: accept: error:14090086:SSL > routines:SSL3_GET_SERVER_CERTIFICATE:certificate verify failed I just bothered to chase this a bit, looks like I'm missing a file at my end. http://www.openldap.org/lists/openldap-devel/200006/msg00064.html I need to make time to figure out why and fix it.
or age histogram? It's not the mailbox that's broken, it's me...
Write a good one with some nice facts and [pointers to] instructions on how to disable them.
ip-up.d/leafnode: hole with funny messageid names? outline-level in file local variables?! ssh with one-command auth key: can you port forward/backward? sauce, can it do TLS with Exim? (tls is best just for end-user connection, I'm not pretending it could replace end-to-end)
tail-mail.pl vs. logprn http://www.coker.com.au/logtools/
Wouldn't it be cool if ... there was a module that could look at the rate stuff is going to STDOUT, and work out from this whether autoflush is likely to be useful. If you're sluicing through huge chunks of stuff, then it will only slow you down. OTOH if you're tail-f'ing a file it could be important.
Find them in Debian. Are they against policy? TODO: insert url here
In no particular order When I get round to it If it still looks like it needs doing when I look more closely
Empty update file should remove the zone Something to list zones? or just use axfr?
Fix it & test it both ways. Try NS & Moz with split version: do they work, do they reformat on save? Use it to unify bookmarks.html! Link into CVS
unified modem control & status project
userv output, script calls and/or piped output drive from system-wide UDP listener or the broadcaster itself defer to CGI interface for back-calls
Maybe the status broadcaster should keep packet state and data for the scrolling usage graphs. Plus a GD interface, so the CGI can draw pics. Maybe an applet would be better...
listener for UDP-status broadcast UDP to piggy
contrib to diald: http://sourceforge.net/projects/diald/
chgrp dialout ; set(e)uid nobody
chown root.dip $1 ; chmod 04750 $1
should probably be more configurable access perms by calling group/id, and per command
allow instant mails to local users
run a oneliner automatically on terminal 6
exim filter by subject =~ m{^for mca/[hw]:}i
hook it from xscreensaver password box
Detect when you're on a console or remote login (and therefore unable to access the X server on which your one almighty emacs session is running) and do a plain "emacs -nw" for quick and dirty. Hmm, except we can't know which X server(s) the esrv socket is connected to.. Can we give it a flag to stop it setting up the Server? Presumably elisp is required in the .emacs, to solve this one. (Detect existing gnuserver and offer to kill it?)
file_todo=~/synced/troop-relapse/linkdir/cvs-PerlUnit/src/Test-Unit/doc/consensus.txt
file_todo_branch={TODO lists}{Matthew}
http://www.jwz.org/xscreensaver/
Add an "unplot oldest" mode, to replace "clear screen and start again" mode. If each flame's parameters (or even point list) are stored, you can take the oldest and replot in black any pixel that has not yet been overwritten by a newer flame.
Is there some reason why you can't increase the density of points? Maybe it's a one-point-per-iteration thing.
(Hmm, it obviously does more than the screensaver shows, considering the CGI version) grokking session first... add log of fetches
because it currently makes a very dull screensaver, if you're on a slow link. cached data, to pick up from last time grovel in Netscape cache? Potentially dangerous!
...like the ones you get when there's a hair stuck in the mouse's eye. Consider a rectangle of x,y which centres on the pointer. If the pointer leaves the rectangle, there's activity. The rectangle could follow the pointer with a timelag or some damping. Once the display is locked, less damping/smaller lag and bigger rectangle
activity on a console tty activity seen by an xscreensaver or terminal (via finger) on an adjacent machine
by UDP, CORBA, <insert buzzword here>, or just make a pty and wiggle that.
eg. pass mouse wheel events to xmms, clicks to the pause button in xmms ...and don't consider these as "activity"
http://www.ignavus.net/software.html
can it scroll? (probably only with a beast gfx card?) or ripple down and start again at the top, with some blank between
List subjects in reverse order, or whatever, like Winbiff Interact sensibly with xscreensaver's notion of "active" Seealso cronbiff
In a .rc file or something? --font="-misc-fixed-medium-r-normal-*-*-200-*-*-c-*-iso8859-8" --color=green --shadow=2
Add or suggest mods. Hmm, it has probably moved on a bit since I got it.
if possible...
It seems sensible to keep them separate from implementation, but this would break code random_network would fit, but it's an instance method too (simple_)?lol_to_tree is a class method importer
are instance methods, how could they go in an exporters module? tree_to_(simple_)?lol(_notation)? draw_ascii_tree
which would be handy (?) for outline2dotty...? conversion to XML-oid and printing is handy, if only for unit testing
Compare http://www.comp.mq.edu.au/~asloane/software/otl2html/ , which generates several pages & does hyperlinks between them
This suggests a shared lib with the DataDumperBrowser thing Perhaps an LWP based fetcher, but that breaks the ACL and causes Safe problems
Use either a Faqomatic/wiki style or that similar CPAN module. Problem is they use a lot of vertical space.
<ul><ol><dl> show these attribs (in name, text block or <small>after name</small>?) heading'ed (instead/as well?) contents are list/pre/flowed ACL props
Scan for /\bseealso\s+(\S+)/i Hyperlink all $1 in a loop with each other Best done with a lookahead walk_down pass? Make permanently bookmarkable #names
Needs better output filtering
read the file list from old carpets and note items that have been removed - allow reprieve of anything with this name
--quiet is your friend allow multiple or even recursive scans recursive would help with the bundling lark
for stuff over (8 weeks) old, remove the [relative age] part and bundle into the current 8-week directory, after unbundling anything inside that
give default username by RP record
Patch to allow a FSF chmod -style symbolic mode The format of a symbolic mode is `[ugoa...][[+-=][rwxXstugo...]...][,...]'. Multiple symbolic operations can be given, separated by commas. to be applied to everything when the "-p" (permissions) flag is set. This would allow a sane solution (eg. 'g=o') when uploading files from a 'umask 02' zone to a 'umask 022' zone. A brief search for a library that deals with these things doesn't show anything obvious. Of course it's a small job, but these things should be done right and done once.
http://crit.org/http://www.t8o.org/~mca1001/crit/backlinkers.html This lot could fit in with the other idea I had. It's all on paper and in my head at the moment. *sigh*
http://www.camram.org/
- Agree something with Lapo about where the pieces live. Code fork is bad. - Consider barging in on HashCash.java looking for optimisations. Could at least profile it, or benchmark on a newer JVM. - Colours etc. if I ever manage to figure the difference between 200 OK and 500 Whoopsie?
Unload the source code somewhere else, and hyperlink it (GPL: needs to be available to the user when they get the binary). Cut out the applet we're not using. It can go with the source version, and is accessible from lapo.it anyway.
Javadocs would be nice. There is dead code to delete, and live code to un-grim.
seems like a lot of effort for one prog, but there will be more
- Check the token when it comes back - Paste an example of a jail message from esj's system into the cgi or docs
The CGI page needs enough words to explain itself briefly, but the bulk of the docs should be going on other pages.
eek
http://www.cypherspace.org/~adam/hashcash/ and a recent message to ukcrypto from Ben Laurie
what's the max. damage possible if v1 just doesn't support bcc'd recipients? the guy on the other end will often enough get valid mail with no hashcash anyway...
There's a pure perl smtpd in Debian, but then exim has perl hooks anyway so why would I care? People were talking about TMDA, http://software.libertine.org/tmda/faq.html
Queue a number, get result back, (eg. NO DIALTONE), try again in 30 sec... Obviously this is going to require some pre-authorisation, but the facility may be handy for those who wish to script a notification of some sort, from a machine which has 'net but no modem.
Simple form for "getting Matthew's email address". You enter your own email address, and the CGI script generates HMAC(password, recipient) + "+" + my_initials + "@" + maildomain Hmm, but then verification gets messy if the recipient address isn't available. You can store a list of tokens issued, but then they might as well be random cookies. Another neat trick would be to send a "please reply to this message" to the address given by the CGI user. They can then reply, and you can check in-reply-to, subject or envelope recipient details.
or find a prog that does this already? I bet there's an applet...
Looks like I want to grow the whole set. cvspublish should probably join the family.
How can I keep a work-mirror of repos-toy up to date but incomplete? Can I propagate changes back again, but with filtering?
I've written a horrible 'cvs2config.pl' but it just maintains a symlink farm. The newish Debian package 'cvs-conf' might be better. Couldn't be worse. 8-)
Check that all files in the cvs directory are pointed to from the real /etc. What other checks do I need? If the symlink has been replaced with a file, that probably means we need to take the file to our (non-root) work directory and hack on it, then do a checkout and symlink. Some things link the files (apache has links to other stuff), others link the directory (bind and exim are more sane).
This probably needs to store user/group/permission status too, at least for /etc and where the script runs as root
There exist Debian-cloning scripts already... (what was I planning on doing here anyway? *shrug*)
(do they also exist already?) Explain everything $CVSROOT/**/*,v and the admin files Provide a script to explain a working directory (hmm, but that's for users anyway...)
cvs-undosify does this, but it's a bit grim
Look for missing modules Look for differences in repos/root/user info Spot changes in sticky status Scan .cvsignore files & summarise
or similar. Monthly job? Can this integrate with the javadoc smell finder? Probably not.
Find & check out all HTML/XML Ensure they have a DTD Validate them (cron job => emails results somewhere, but see cronalike) This might be better done as a web crawler, eg. for JSP or CGI progs
Filter errors from scripts such as above with a 'cvs annotate', figure out who the culprit was and direct the email that way. Not likely to work for CGI output.
wrapper(?), tied to checkin below a module goes places and pulls back updates optional updates for things, eg. 'dns-update && ndc reload' after zone changes Could take a port-forward with it when it goes (to CVS down), but this doesn't work if you send two at a time (unless to pick random ports and then screw with CVS at the other end?), and it makes any other operations done from the other end much harder.
Since I have full history on the file (indeed the whole repository), I can plot any properties I like, against time. Better still, I can write the program any time I like, and still get the same results! So, I'll do this later. Much later.
[Will drag stuff from section below if it's still remotely sane.] I want to make encrypted time-stamped copies of some (all?) cvs files (esp. the todo list & programs) for "prior art" value. The prior-art-o-matic program is coming to life as a suite of little utilities. This seems to be a useful route.
Originals need storing, unless you can guarantee that they'll be available in their original form. - emails, permanent: store encrypted to self, header & body separately encrypted, and then you may delete the original. Permanent copy should go in CVS or RCS to retain the oldies. - emails, temporary: just keep the one in the folder, but make sure the headers are digested separately each. May want to grovel the binary attachments etc. as files, ick. This seems much more sane, especially if you archive your mail anyway. - cvs content, permanent: this involves keeping a copy of either all the versions (nicely diffed), or each version on a branch with an extra file per branch, or track the raw CVS file. - cvs content, temporary: just check out each version in turn. You then might want to keep a copy of the 'cvs log' output, but this could be done for the whole repos.
You need a name for the thing you're storing so you can get it back later when you need the original of that date. For files such as CVS content or maybe an mbox of postponed mail, - name is easy, just the filename and offset/revision etc. - you want to store diffs not whole copies per change For general emails, you may need to store by message-id and some other label (in the case of duplicates). Much better though, would be kill the dups first. 8-}
For when you don't want to bother dragging in each version of a file, perhaps? Anyway, I dragged in my CVS repository as an import into another repository. Lessons learned - consider fixing the various wrappers and ignores in the new repos - rename CVSROOT to something else - rename the source,v files first otherwise confusion will reign when you come to store only the latest revision, or something. Ran with "cvs import -d -kb -I \! . mistuff REPOSFILES WDL1" This is a bit icky; running regular automated imports and expecting to cope with removed files or conflicts is not great. Maybe this is silly. Maybe I should just take a backup (a back-what?) and stamp a digest of that. *sigh* Actually no it's pointless and stupid. You can't prove anything much about early versions of the file without revealing (probably) large chunks of the later content. The only benefit of this is direct info on the tags and other metadata. This should probably be taken from "cvs log" and .. there's other gunk in that header. I want it.
bidirectional secrecy (assume stamped work is effectively public) selective revelation, to some granularity this allows you to reveal _a_portion of _a_ file at _a_ point in time
encrypt the lot to a long term public key of your own, stamp that, and rely on revealing the underlying symmetric key if you need to.
plaintext list of md5s per paragraph for each file when it changes rely on CVS for full text update paramd5s to comprehend paragraph breaks in vanilla outline-mode
pick a session secret, S use a sequence of symmetric cipher keys H(S+0), H(S+1), ... encrypt each paragraph (you can probably get away without encoding length, so no expansion. May be able to compress each block separately, so some compression) stamp the ciphertext encrypt to your long term public key the secret S for fixed-size blocks this might make a neat gpg addon?
This would be an "anti-rootkit" port scan. Bind to each port in turn, check for unexpected failures Not spoofable by packet filters or whatnot. cf. nwatch, packaged for Debian. Pretty cool.
a cron job that knows which mail folders are important/pending, which are just archive storage and which are junkmail in jail, and sends a periodic summary?
Also should be able to summarise (count on some criteria?) number of items
gnome (or other) menu items to ssh to various hosts generated from a sensible config file open terminal and ssh ssh -X and run terminal at the other end
Hey you could use a menu generator to list the mutt folders too. Not exactly interactive though.
Dave Scott's idea: > [...] someone ssh's to your box with X forwarding enabled. [...] > BTWBTW it would be a fun thing to spot these insecure forwards and > xmessage 'you really shouldnt do that'. That's a _fine_ plan. Presumably you could stick the thing in the .bashrc/profile, which avoids the need for a r00t daemon. Also the facility to touch a file like toy:.ssh/xhole.I-know-and-dont-care:relapse.granta.internal Same goes for auth forwards. I believe you can detect both from the environment, on the assumption that nobody's going to try to stealth them. I'll paste this into my 'todo' list, and you can watch it appear in the vapourware queue on the next update? 9-)
I'm going to need "is a ToDo item" and percentage complete pretty soon!
http://ftp.ru.xemacs.org/pub/emacs/emacs-lisp-packages/distribution/misc/ http://ftp.codefactory.se/pub/people/daniel/elisp/todoo.el
That doesn't work for grafts between levels
'cos using Graph::* would be a Bad Thing, right? (why?)
There must surely be tools now, but if there weren't I would (teehee) write something with which you could mark classes of package as interesting, never interesting, installed on a temporary basis, or whatever. The ability to annotate packages by what you've read about them may be handy.
"Things you need to know if you didn't bother reading the rest of the docs" or "Things I found out the hard way" >> I've still not read the docs! Oh I'm such a bad man! 9-) If you're doing multisession, you need the "multisession" option but not "disable fixate". That's for special audo stuff, and if you do this you'll have to fix the session yourself later. Normal CD drives will fail to read the disk with this option. Multisession bootable CDs need to either change the catalog filename (perhaps bump a serial number on it?). Don't forget to include another copy of the boot image. Alternatively you can switch "bootable CD" off, and it will do just that - the previous boot image doesn't seem to stick around [hrm, maybe this laptop is just a bit funny?]. Note that just changing (or adding) a boot image will still cost you 22 or 13 meg, according to the "multisession" button's tooltip. Liberal use of "dummy mode" is helpful if you're a dummy. Yes I know it's tedious. Use it for the second session! It takes quite a while to "blank" a CD-RW disc. The comment on "Eject target when done" says some drives need it. I got a feeling my LG-GCC DVD/CD-RW drive didn't like it, but that was when I was confused about the "Disable fixate" option.
>> Check out Gibraltar... When you're writing a proper root filesystem, don't forget to turn off "Anonymous Rockridge" or things like ssh will be very grumpy about permissions. Putting your favourite host keys in /etc/ssh/ssh_known_hosts might be a good plan too. The pcmcia cargmgr requires writable /tmp, as do many things. You can get this with a tmpfs or shmfs [what's the difference?], or if really desperate you can make a filesystem on /dev/ram and mount that.
Starting with a BIND which serves an internal domain plus external cache (ie. cache on eth0 not loopback, but still to your local net only). I believe this is a common setup... If you're on the ball you may have BIND bound only to eth0, to make it available only to the gateway host and the other local machines. Otherwise it will be serving your internal domains and security weaknesses to the world. As outlined already in the various FAQs: you need dnscache on UDP eth0:53 so the other inhabitants of the local net can send recursive queries to it. you need tinydns on UDP 127.x.0.1:53 because it can't go on eth0 and because it probably shouldn't be on your externally visible IP address. 'x' is often 0 or 53. So you ask, "surely this means that local machines will no longer get authoritative answers to queries in the local domain?". [and THIS IS WHERE I'M FUZZY, please correct me if I'm wrong!] Yes, but they never could get authoritative answers for any other domain anyway, because of the cache. This will only be a problem [?] if you choose to start using dnscache on machines on your local net, perhaps a laptop which at other times lives on public ethernet somewhere. dnscache will only accept authoritative answers, to avoid cache poisoning. (but it has a forward-to-cache mode, so this can't be true?) I think the short answer is that using dnscache on other machines is pointless, but if you want to do that then you would need to secondary the internal domains on that machine too, by using axfr or some rcp-sync method. ARGH! dnscache binds the TCP port too, so you can't run axfrdns on eth0. Oh, and you'd need to tunnel DNS requests through your firewall, perhaps with NAT or masquerading.
http://www.alcove-labs.org/en/software/sonypi http://www.alcove-labs.org/en/software/meye
OK, I've got to find/write one first Check out also rcsplot (under CVS tools)
http://www.debian.org/ ( dpkg --get-selections | grep -v deinstall | cut -f 1 cat pkg ) | sort | uniq > pkg2 grep -v /dev/ /var/log/setuid.today | cut -c 77- | xargs dpkg -S | sort
allow user anacrontabs, like the standard user crontabs Maybe this goes along with hooks for boot time?
doesn't spot -*- mode: foo -*- hints in text files (I want to note outline-mode files)
locks to /var/lock/cvs and set group on that directory then doc it
.gz .zip .exe .dcr .dxr .gif .jpg &c. Hit on the system MIME-type list? Is this an unnecessarily large and inaccurate job?
Dies silently if leasefile & pidfile can't be written? Didn't check the exit status or verbose option
suggest "--dd" for double density disks
Adds access controls to /etc/apache/access.conf instead of httpd.conf That's the old way of doing it (and confused me, since I now neglect that file)
dctrl writes to fifo, diald only deletes pipes possible problems with system default /etc/ppp/options, too?
Netscape or Windowmaker? Something went into "drag a window" mode, and wouldn't leave. The keyboard was Grabbed by something, so even after killing NS & WM the X server was still pretty useless. Bug should also go on xscreensaver - it could tell the keyboard was already grabbed when I tried to unlock, but it still locked the server.
Desired=Unknown/Install/Remove/Purge/Hold | Status=Not/Installed/Config-files/Unpacked/Failed-config/Half-installed |/ Err?=(none)/Hold/Reinst-required/X=both-problems (Status,Err: uppercase=bad) ||/ Name Version Description +++-============================-============================-======================================================================== ii xscreensaver 3.31-2 Automatic screensaver for X ii navigator 4.76-1 Meta package that depends on other packages ii wmaker 0.64.0-5 NeXTSTEP-like window manager for X ii xserver-xfree86 4.0.3-3 the XFree86 X server ii libc6 2.2.2-4 GNU C Library: Shared libraries and Timezone data
The thing called /usr/sbin/dnsupdate.pl is for dhcp-dns
Apparently the thing which sets the repository is the working
directory, $FROM. If you have a CVS module with g=rwxs to the dnsadmin
group you can use this instead.
cd /var/lib/dnscvsutil/domains
# pay attention to file perms!
ls -l . CVS
# this old state points at the package's repos
rm -r CVS
cvs -d $CVSROOT co -d . your-dns-module
# this should happen already, since domains dir is setgid
chown -R root.dnsadmin CVS
# this is what we had, but it looks too restrictive?
chmod 0755 CVS ; chmod 0644 CVS/*
# hmm, thought so (maybe a default root umask problem?).
# Oh, and the group doesn't stick. Let's try
chmod g+ws CVS
You can probably prune /var/lib/dnscvsutil/cvs too, but I'm tolerant
of cruft.
Need a { warn "You forgot to set the origin" unless defined
$vars{origin} } at line 174 before 'return <<EOF'. Maybe fix the
default thing too.
What's going on? 1.10.8-1 works in local mode 1.11-? and 1.11.1-? seem less happy, possibly as repository servers
If you've got a (pair of) speaker(s) in each room, then you could add a mic too (and make sure you stay in control of it!) Then you can use a pocket DTMF as a remote control. * = read out current menu? # = abort command?
Should be easy if you wire into the back of an ext voice modem
Although the paper has been publicly visible for several months, it needs diagrams etc. to be "finished". This is one "to do" item. There are more "to do" items listed as "Ideas and further work". I'll copy some into this list so they can be forgotten without prejudice.
A simple daemon would be good. For performance on old hardware the bottom network layer probably needs to be in C, but it can then hand off results to something else. Perl seems likely if I write it. 8-) I can avoid the "piling up stale data in the TX queue" problem by rate-limiting to very slightly under the measured throughput. That seems like a neat solution, unless the TX queue fill level is available in userspace.
In the absence of any beam to control, hardware to control it or interface to that hardware, it seems necessary to write something to emulate one for testing purposes. This sounds like a very unrewarding task, but I mention it for completeness.
Derek Weston (UPN Laser Transceiver designer, http://www.alphalink.com.au/~derekw/ ) kindly emailed me to say he made such hardware (loudspeaker-mounted mirror) some time ago when he had this same idea. I should pick his brain before (if) I build my own.
http://www.svtehs.com/optolink.htm
(yeah right)
hmm, seems this problem has been solved already, at least to some extent http://www.cs.uoregon.edu/research/paracomp/anym/
give it a list of words you're after, marking alternatives and order a list of valid words a thesaurus? some CPU time can I find something faster than depth-search? can I make a soundex thing to sort pronounceable non-words from rubbish?
http://sammy.net/~sammy/hacks/ I had it working once, but now it doesn't. Not sure what I did to break it. 8-( Suspect the change of kernel (2.4.x to 2.2.x) may be involved. I get "connect: Invalid argument" when it scans the remote (inside the firewall) hosts The UDP socket in question is bound to the address of the _sender_ of the packet that triggered the fork, but any old port number; the connect to an address on a different interface fails. Maybe I could avoid the connect and just do sendto() or whatever it's called? Somehow I don't think it's going to be happy. Dave found a page that suggests ipmasqadm portfw -a -P udp -L 0.0.0.0 6112 -R 10.0.0.5 6112 in addition to the existing masquerading of tcp and udp, ipchains -A forward -i $IFC -j MASQ --dport 6112:6119 -p tcp -s localnet/24 ipchains -A forward -i $IFC -j MASQ --dport 6112:6119 -p udp -s localnet/24 It works nicely for multi players, but won't support multi players behind one cable modem. That's not my problem, so I file this lot under "document and/or investigate later".
Grabbed from some web page somewhere, "I use Starcraft behind a 2.4.0 (RH7) masq firewall using netfilter/nat/iptables and it works fine by default. iptables -t nat -A POSTROUTING -o ppp0 -j MASQUERADE (or something very close) Under redhat 6.2, stock kernel, it also worked. Under 7.0 stock kernel, I had to echo 1 to that /proc file you mention below.
http://www.grantadesign.com/
fix MySQL privs: _ --> \_ for DBs
Now exists and is called dns2dhcp.pl See also http://dotat.at/progs/zonemaint
Finish trumpton.internal, then do the reverse zones Hmm, where are the reverse zones now?
Burn a CD. Always the easiest way.
Not convinced it's fast enough
More/same on my int home page...
Easily done with standard parts, once I've got clustered 'mon' operating [huh?] NFS is not necessarily my friend... "smbclient //box/share -c du", seems to permit anon login. How hard would it be to make a "fill rate" detector?
Get properties via LinkState modules (seealso diald/pon) Not blocked Last connected within (sensible time period, probably depending on time of day) Check outstanding mail Compare/extend the mrtg-linklive.pl thing
slug some mail round and see it go
Anything on the queue more than a certain age? More than a certain number of items?
Run an exim config test (surely the admin wouldn't leave it broken?) 9-) Check up & down are almost the same
Check a set of machines have the same time, +/- five minutes, using any available time protocol. (How do we get at SMB time?)
http://www.kernel.org/software/mon/
Poke the monitor at poll-interval It checks what needs doing, keeps state on last update for each thing (doh! this is already implemented. Maybe it's time to use it.)
A whole 'mon' process or some subset? Run a socket connection one way or the other for status? Or just have a whole separate monitor and put the two in a frameset? (doh! again - mon has a socket, so we need a process to do dependency)
Seems a little OTT, and not actually useful just now...
Hmm, isn't this the wrong way around? Exim has hooks to get its mail config via LDAP... http://www.exim.org/
/etc/aliases forward lookup Probably not .forward files, since that's covered below
/etc/aliases reverse lookup
/etc/aliases for funnies .forward for current destination
Last read (atime) Most recent mail (mtime)
Add remaining inverse gateway functionality:
Install an existing program? http://k12linux.mesd.k12.or.us/ldap/