Recent blog entries for splork

Places where policy doesn't belong...

A former boss asked on a mailing list if anyone knew how various OSes chose their source ports for TCP and UDP connections. we got answers. i decided to look at the code and supply an answer for linux 2.4.18 and found the following default source port ranges:

  If the system has >=128mb of ram:
     32768-61000
  If the system has 64-127mb of ram:
     1024-4999
  If the system has 32-63mb of ram:
     2048-4999
  If the system has <32mb ram:
     3072-4999

Sigh. Why? This is a policy decision. If someone has a system with low memory and thinks that this will actually make a difference to their system: Run sysctl in a startup script! sheesh.

Note also that the IANA standard for source ports says to pick a number between 49152 - 65535.

Maybe I'll write up a patch, post it and be laughed at. Or maybe i'll be lazy...

daniels i don't keep up with the lkml these days but I went looking when you mentioned bitkeeper annoyances. heh. i couldn't agree more. larry mcvoy is an asshole. i don't care how good bitkeeper may be; its attitudes like his that make bitkeeper pointless to use.

yippee. i like the new posting indicator and trust metric score displays.

*please* can't we simply prevent duplicate story postings in the advogato software! computers are good at that kind of thing.

Joined the EFF using paypal today with the "DMCA" listed as the reason. (thanks for the reminder in someone elses diary, I had been meaning to do that).

Also sent a suggestion to paypal: I would like a way to show proof of payment to third parties. (so that Alice can prove to prove to Cathy that she paid Bob).

It could be done either requiring Cathy to be a paypal user (they would like that I'm sure) or alternatively to anyone not even being a user by making the page a "one-shot" that after the viewer clicks a "confirm that i have read the proof" button it goes away [so that one proof couldn't be used multiple times].

I suggested it for the purpose of selling items saying "proceeds will be donated to the EFF" (or charity of choice that accepts paypal). That way buyers could be sent proof of the truth in that statement.

The alternative "fun" use would be for thugs to get proof from you that you made your weekly blackmail payment to the thug's boss.

vab: take a look a Mojo Nation for your free software distribution p2p network idea. Using a p2p network such as mojonation or freenet to distribute packages would be excellent. (can anyone drool as they envision apt-get using a p2p network to get packages instead of silly ftp/http mirrors)

Mojonation and freenet both have the advantage of content being addressed by secure hashes. The only thing that would need to be signed (if you actually implementing signing) are the indexes that list what the hash ids needed to get the packages are.

yippee, its been a while in the making - we "shipped" a new version 0.970 of mojonation today.

i went for a bike ride in the morning after dropping my car off for its regular service & smog. i (foolishly:) volunteered to go on a 40 mile morning ride with a friend in a couple days, i need to build up my endurance again.

xgavin pointed out a nice article on irc: Robert C Martin on expressiveness and language choice. Of course I call it good because it pretty much sums up a lot of what I believe. :)

PlayStation 2's are evil, they take up too much time. My roommate won one from a radio station last week. Oh darn. :)

Released a new version of the python interface to BerkeleyDB 3.1 today. It now works with python 1.5.2 and 1.6 out of the box. Get py-bsddb3 2.2.0 here.

I am very happy with the python distutils package. I just converted py-bsddb3 to use it. I have noticed the roughness around the edges in terms of unimplemented or undocumented features but the code is easy enough to read to quickly figure out what it can currently do. (though I do have to say I prefer undocumented features to unimplemented but documented features!)

(sigh) I should join the python-dev mailing list. For a while py-bsddb3 was being considered for addition to the next python release but I haven't been able to follow up on that. Someone even started creating an un-swigified version that I should try and make time to follow up on.

busy busy busy...