:: 31-Mar-1999 05:20 (Wednesday) ::

I’ve managed to get several portions of the website finally cleaned up
and updated this weekend. Most visibly this includes our old source
code distribution page, now at http://www.distributed.net/source/
instead of /cores/ as it was before. There will probably be an
updated bundling of our public source sometime in the near future,
though I can’t give you a precise estimate of when.

Also made more prominent links to some of the more obscure pages on
our site. The unmaintained news pages are now gone and replaced by
our single History & Timeline page.

I have also made available some of the vectorized Encapsulated
PostScript (EPS) images of a few of my favorite cartoon cow pictures.
These EPS versions can be enlarged to extremely large resolutions
without pixelation effects. The original low-res bitmaps from which
those were derived contributed to many of our graphics, including the
client application icon used on the Windows, MacOS, OS/2, and RISC OS
GUI clients. You can grab them from http://www.distributed.net/logos/
These might be useful to people planning on submitting poster entries
in our design contest (but you’ll still need to be creative to win).
See http://www.distributed.net/poster.html for details.

In general, things have been rather busy for me since I am a senior
and will be graduating from Harvey Mudd College this May and have been
fussing with job choices and such. Furthermore, I’ll be traveling to
Eindhoven, The Netherlands next Monday for a full week with my
teammates to compete in the ACM International Finals
http://acm.baylor.edu/acmicpc/, so I’ve been tied up with preparations
for that as well.

I’m hoping to be able to get the next build of the personal proxy
released before I leave, otherwise it’ll probably get released when I
come back. There are a number of new features that I’m sure many of
you will greatly welcome.

:: 30-Mar-1999 19:02 (Tuesday) ::

Team joining code, stage 1 is in progress right now. One of the
components of the team joining code that’s necessary is a query that
will be run each night during the daily update.

The purpose of this query is to retro-actively assign blocks to a team if
that participant has just joined a team and has unallocated blocks in the
database. For example:

A participant fires up a client for the first time on day 1. Days 2 and
3 pass as the client completes more blocks, and on day 4 that participant
decides that it’s come time for them to join a team.

This query identifies the blocks (currently not assigned to a team) from
days 1, 2, and 3 and retroactively allocates them. So, when the
participant finally joins their team, all their _unallocated_ past
activity will go to that team.

This differs from the situation where a person changes from one team to
another. Once blocks have been assigned to a team, that assignment is
permanent.

Why am I telling you all of this? Well, for two reasons. First, It’s
been quite some time since we’ve had a statsbox that did this, so it’s
quite likely that there are people out there who are unfamiliar with how
this works.

Secondly, and far more importantly, I’m about to run this query for the
first time. And there’s a bunch of blocks in the database that will be
affected by this. See, towards the end of the lifespan of the original
statsbox, this query would quite frequently fail to run properly.
There’s actually a non-trivial segment of blocks in the database that
aren’t allocated to a team properly. All those blocks are about to get
assigned in one mad rush.

The result of this will be some pretty strange team ranking issues in
tonight’s stats run. It’s conceivable that dead teams will rise in the
rankings, it’s quite likely that some teams will see their “time working”
date roll back some. I thought it would be remarkably preemptive of me
to warn everyone ahead of time.

:: 30-Mar-1999 16:03 (Tuesday) ::

In case anyone hasn’t noticed, the team summaries are now linked into
the team listings. Yes I do know about the bug that reports a team as
unknown, even when it is in the listing. I’ll get that fixed as soon
as I can (silly real life :P).

Here is my “priority list”:
1: Participant listing
2: Team information editing
3: Team history listing/graphs

:: 29-Mar-1999 20:42 (Monday) ::

Thanks to dbaker, the password mailer on stats is now operational. And
the first bit of the participant configuration editor as well. It’s
still not possible to join a team, but we’re /very/ close to that
being possible.

:: 29-Mar-1999 16:33 (Monday) ::

A few minor things. I’ve been mostly swamped with hfdirect work the past
few weeks so stats development has been slower than I’d like.

o I’ve had a handful of reports of underreporting on stats. It’s not
widespread, but enough that it’s also obviously not user error. So far,
in every case the numbers in stats have matched the numbers in the
keymaster logfile. If there is a problem, it’s upstream from stats.

As always, if you suspect there is a problem with your clients / work
being lost, the first thing to do is enable logging everywhere you can.
Detailed logs will allow us to trace your blocks through the keyserver
network and identify exactly what is happening.

o http://rc5stats.distributed.net/cgi/emsummary.idc?em=nugget@slacker.com
works once again. This is a legacy url from the old stats box. This
means that the “view my stats” option in the win32 gui client should once
again take you to the appropriate page.

:: 28-Mar-1999 19:41 (Sunday) ::

It has already been 2 months since I took over regional support
coordination from Decibel at the end of January, so I guess that it’s
about time I updated my plan.

I’ve been doing a lot of ‘behind the scene’ work to make sure that our
ever growing team of regional representatives is ready to provide you
with fast and reliable help. As of today, 45 countries offer regional
support for distributed.net.

I have also been working on implementing a multilingual website
management tool that will allow our regional support staff to
translate web pages for us. These pages will be accessible through
your browser if you set the language priorities properly (ie: set a
higher priority for French than for English, for example, and pages
that are available in French will be displayed instead of the default
English pages). More information for Netscape users is available
http://help.netscape.com/products/client/communicator/nethelp/Navigatr/nav.htm#PREFERENCES_NAVIGATOR_LANGUAGES.

I still have to clean up the code and fix minor things before a small
group of regional representative beta testers can actually start using
a functional version of the translation tool. On the whole, I’ve been
making significant progress on this project and beta testing is only a
couple of weeks away.

On a totally unrelated note, I set up a search engine for the d.net
website. You can use it at http://n0cgi.distributed.net/cgi/dsearch.cgi.
A query form will soon be added to the navigation bar of the
http://www.distributed.net/ page.

:: 28-Mar-1999 03:59 (Sunday) ::

Thursday evening I lost control of my car when I was at ~150 Km/h
(>90mph) on a small road going though a forest. Fortunately I was
alone, in the car and on the road. Everything was done in a split
second. I probably lost conscience when the car hit the first tree.
When I waked up, I get out of the car through the window (I couldn’t
use the doors anymore). Then I stopped someone on the road and he
drived me to the hospital.
I really realized the importance of the crash when I saw my car at the
mechanic the next day. Everything was broken outside and inside the
car. Even my seat was broken ! There was blood everywhere.

I’ve seven stitches on the face, but beside looking a bit like
Frankenstein’s creature, everything is ok. I can’t believe I’m still
alive and (mostly) healthy…

Morale of the story : alcohol + speed = crash.

If you’re used to drive far beyond the speed limit while under
alcohol, sooner or later you will face the death. You might be (very)
lucky like I was. Or you might not.

:: 26-Mar-1999 02:53 (Friday) ::

http://www.distributed.net/bugs is open for business

this is the Distributed net version of Bugzilla – the Mozilla project
bug reporting system

:: 23-Mar-1999 05:43 (Tuesday) ::

As you’ve most likely noticed, I’ve been able to massage the stats database
back into sanity and things are running smoothly again. There’s code in
place which should detect if this runaway index problem strikes again which
will make corrective measures go much faster.

Anyone out there who was assigned an ID in the 5-6 million range has a new,
more reasonable ID in the 170,000 neighborhood.

Many thanks to Bradley Langhorst and wayne4@bellsouth.net for the tips
on Sybase identity field issues. While we still don’t know exactly what
is causing the behavior for certain, I think we’ve got a few good
working theories now.

More news as it happens… :)

:: 20-Mar-1999 06:17 (Saturday) ::

The stats update this evening failed. Apparently a few days ago we got
hit by a bug that has happened once before. When adding in the brand-new
emails into the participants table, for some reason the identity column
(that becomes your ID number) goes nuts and starts incrementing wildly.
Newly-assigned ID numbers should be in the 160000 range, but recent new
emails have been assigned IDs in excess of 10000000. This causes an
overflow when updating the master tables because the primary
RC5_64_master table only has 7 digits allocated for the id field.

Correcting the problem is not difficult, but it’s a lengthy, manual
process and I don’t feel comfortable going through it when I’m tired.
I’ll have to manually re-assign the bad ID values and then manually
correct the ID numbers in the master table.

I’m still not certain what causes this to occur. At the very least I’ll
want to add code to the daily update routine that will detect if this
happens again so the error can be corrected in a more timely manner.

I did re-run an “update statistics” on the participants table today,
which is not something I normally do. It’s possible that this activity
is related to the error.

Sorry that the 19-Mar stats are delayed. All the data are fine and no
blocks have been lost. Look for the numbers to be re-run sometime
tomorrow.

Thanks for your patience.