Older blog entries for ladypine (starting at number 71)

Orkut stores email adresses and uses them!

Orkut is a community service, to which one can join by invitation only. I joined several monthes ago, and invited some of my friends. I deleted from the friends list the emails of those who did not wish to join orkut.

In the past few days, several of those deleted emails got an invitation from orkut (sent as if it was sent from me) again. Deleted emails, wrong (bouncing) emails, emails of friends who have not yet responded - all those were sent again.

Orkut made me appear like a spammer. And worse - big brother is storing you.

haifux logo contest has entered the voting stage: choosing a voting system

We currently have 14 basic designs on the haifux logo page. I am really impressed with the thought invested in each of them.

As for the voting method, epsalon suggested the Debian voting system. At first it looked really difficult to implement, because it is explained using graphs. Thinking of creating a graph with 14*14 edges is horrible. No chance that it will be a flat graph. But then I read the about the Condorcet voting system, upon which the Debian one is based, and I realized two things:

  • The most intuitive implementation of all those voting methods is a matrix, not a graph
  • The Debian voting system is suitable when you are willing to take "no decision" for an answer, and you want to set the conditions for such an event. If what you really want is to know the option most prefered, then it is the original Condorcet method that is the right one.

And then there is the problem of cycles. These systems are based on relative preferences. I, as many others, had the impression that a system which is based on partial order relations is bad, because you may reach a loop: A>B>C>A. And then what do you do? But as is rightly explained about the Condorcet system, this is not a problem of the system: it simply reflects the real preferences.

Another problem with methods that are based on preferences, is that the data is a matrix and not a vector, and it may be hard to follow the way in which the best option is chosen. Any method in which it is hard to figure out how exactly justice is done is bad, because of the lack of visibility. What is simpler than "the one with most votes takes"?

I hope I solved this by giving lots of examples about the partial ranking, and by suggesting to vote for only one logo, if the voter so chooses.

So, all there is left to do is accumulate the votes over the next two weeks.

I prefer original documentation any time

And so do most people. However, when mulix suggested in private mail to someone to read the original documentation, he was threatened by prosecution!

The following mail was sent to the general openMOSIX mailing list by Moshe Bar:

Date: Thu, 22 Jan 2004 07:18:27 -0500
From: Moshe Bar <moshe@qlusters.com>
X-Mailer: Microsoft Office Outlook, Build 11.0.5510
To: 'Nemeth Lorant' <loci@crandon.sch.bme.hu>,
Subject: RE: [openMosix-general] Load, Big Mac Index
This guy Juli Ben-Yehuda is not an oficial voice within openMosix. He ha so
far only taken the voice of the one who slanders and critizes without
grounds and points people to the wrong direction. A team ot top-lawyers are
waching and documentiong his wrong moves like this one and he and his
friends will be prosecuted in court soon.
Having said that, Mosix has long dropped the comparative resource prizing
scheme and has opted for something that is completely differnet and that is
completely impossible to document.
To anser your questions:
IN openMosix a node (a PE) sends information to all nodes, at random. So a
parituclar node might have gotten the load laevel info from node A only
after some time experired. So node 22 will now first have to contact node A
and ask it if it's trust what is being told that it sload is supposed to be
such and such.
If you have more secific questions, go ahead and ask me. Ignore Muli who has
his own personal agenda to destory the reputation of our project (for some
Kind regards
-----Original Message-----
From: openmosix-general-admin@lists.sourceforge.net
[mailto:openmosix-general-admin@lists.sourceforge.net] On Behalf Of Nemeth
Sent: Thursday, January 22, 2004 7:04 AM
To: openmosix-general@lists.sourceforge.net
Subject: Re: [openMosix-general] Load, Big Mac Index
I couldn't find anything about the definition of load, and Big Mac Index on
www.mosix.org. :( Any other ideas?
The questions:
- how does load in openMosix differ from linux load (it calculates with the
migrated processes also, but any other diffs?) How can the same process
cause biger load on a slower machine
- How Big Mac Index works in openMosix (country=node, goods=resources...over
and undervalued currencies?). What is the meaning of the original terms is
- How can the overhead of management be linear in the count of the number of
nodes if every node collects information of radnomly selected nodes?
(because n/2 * n is in O(n^2) if there isn't any limit of how many nodes can
be choosen randomly to send and recieve status info. from or to)
On Thu, 22 Jan 2004, Muli Ben-Yehuda wrote:
> On Thu, Jan 22, 2004 at 12:27:42AM +0100, Nemeth Lorant wrote:
> >
> > - I've read about Big Mac Index algorithm in Moshe Bar's ,,How
> > openMosix works'', and also checked some docs about it, but I still
> > don't understand how openMosix adopts this algorithm (country=node,
> > goods=resources...over and undervalued currencies???).
> Ignore that crap and read the MOSIX papers (and book). Links on
> http://www.mosix.org.
The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open
Tools Development and Integration See the breadth of Eclipse activity.
February 3-5 in Anaheim, CA.
openMosix-general mailing list
3 Aug 2003 (updated 3 Aug 2003 at 19:28 UTC) »

Finished my impressions of OLS, in Hebrew, for tomorrow's Haifux. Rusty's list of "must-know"s came out very interesting indeed, in some parts. The other parts I knew were good. I just cannot figure out why he pointed watchguard.

August Penguin tomorrow.
Somehow, I am not half as excited as last year. At least the book crossing might work, thought the mind games will not happen, nor selling free software books on discount.

Back from OLS

Just came back from OLS. Ottawa is a beautiful city indeed, though I have not seen much of it. I found many of the lectures hard to understand, and in general I might say the lectures were divided to really good ones and really bad ones.

Mat Porter gave a very good overview of porting, though it was highly technical and I got lost on the way. I wish I could understand it fully- this is how a porting lecture should look like.

Paul Mackerras's lecture had a very smooth bottom line: optimizing based on reading the code only (without profiling) may end up gaining nothing in performance.

Patricia Gaughen was insipring: she made me want to get there and try to solve those problems: for example, make it possible to migrate pages between NUMA nodes, at first withought trying to optimize it. Though at the BOFS somebody said that it is posible to migrate pages, only it turned out inefficient: in that case, that is a very tingelling research problem, of when to migrate pages.

Rik Van Riel of course gave a hit lecture, called Towards an O(1) VM. Although I lost him in the middle (as he said himself, seeing the amount of people in his lecture, he could not expect everyone to know his way in the VM: there are not that many people who fully understand it!), I enjoyed his talk very much.

Meeting people I heard of on lkml, people I heard of from mulix, and hackers from all around the world was an interesting experiance, though not always easy. The standard view of the lounge was of people with their nose up their computer (wireless internet was available freely, and an email garden was established). Making eye contact is a mighty hard job, when people never look at you. The gpg signing was a good ice breaker, but unfortunately I was late with submitting my finger print.

Still, I met benh, lmb, sarnold, zwane and behdad, who were all very fun to be with.

Packing to go to OLS. At least mulix knows who the people around are. I can count two people who have heard my name at all, and two people I expect to recognize. Muli claims that OLS's social side is much like haifux, but when I think of how long it took me to feel comfortable in haifux, it does not cheer me up at all.

I do not feel comfortable about how deep will my understanding go in the talks themselves. Muli has offered that I read about the Linux kernel on the way. I believe it is a bit too late to become a kernel hacker in a day...

On the other hand, the last time I went to a conference I knew almost nothing when I came, but I knew a lot more when I left, even if I did not understand thing right at the time.

I intend to go to OLS. Registered hastily, in the last minute. Out of three conferences I tried to go to, I failed to arrive two. (beurocracy can be very strong). I hope this time will be different, though.


Miming version, not drawing.
I have been trying to gather a group for a pictionary game for about a year now, and finally it happened. We had a lovely evening of people who not necessarily knew each other to begin with, and yet acted and mimed and enjoyed themselves.

And of course, I placed on the tables twice as much food as was needed, I had twice that amount in unopened bottles (where did that bottle opener go?), and I found an equal amount of serving plates (which I had worked hard to prepare) in the fridge, after the guests left. Luckily, I listened to mulix in time, otherwise those vegetables which I left untouched would have been chopped as well.

We tried to have an evening of not-geek jokes, but it seems that most of uswere having troubles in that direction. It is kind of like playing "yes no black white" (the person who says either of these words first loses), where the forbidden words include Linux and computers. We still gave Win98 support at midnight, and still when I asked Muli for some CDs to play, he came with Gentoo-Knoppix. Personally, I prefer listening to Enya.

Portable Programming Lecture in haifux

The portable programming lecture went pretty fine, compared to what I feared. Yes, I did lose my thread of thought several times, and got mixed up with my tongue (saying "backslash backslash" in front of other people while being recorded on video is a nightmare).

About 20 people came, most of which I knew. No masses of people from Tel- Aviv whom I have never seen, and on the other hand- not a lecture for 5 people with no interest whatsoever. Rather surprising, since I did not think people cared about portability at all. After all, most people do not have many machines to run on.

kilmo, mulix and Eli did a great job at helping me keep the discussion under control. This is what it is, actually (as Muli said): Guiding a discussion, rather than giving a lecture.

OK, I crossed the line, now I start thinking about the next lecture.


Maud, my cat, had an operation today. She is the poorest cat I have ever seen, dragging herself around, wearing an Elizabeth collar. She has a piece of plastic sticking out of her shaved skin, to allow for the fluid to pour from her. She has not eatten for a week, and she has lost about 1 Kg of her weight.

62 older entries...

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!