Should we, could we, and how do we encourage slow players to play more quickly or leave the tournaments on the site. If it weren't for a small number of recurring offenders tournaments would fly round. Would it be possible to filter them into a separate division or tournament?

Would be fine with me.

I think the time out rules are fine. I am not in favour of creating new rules within tournaments.

Two things I would like to see changed which might releaf some pressure:

1. Automatic removal out of championships when player loose by time out. (Note that removal means for next rounds, and that players can sign back in without problem)
2. The time out demon could be programmed that it is triggered back 'awake' when a negative time occurs on a game, as it is now a failure of the bot has to be triggered manually by richard, which on occasion happened only a week or so later (when richard does log in with his mobile he can't trigger the demon)

It might be a good idea that whenever someone complains in the forum about loosing games on time, claiming malfunction of the server, after having used apparently all his vacation we keep explaining to them that still they have chosen themselfs to play so close to time out and thus are themselfs the first responsible of the time out.

I agree with #1 with one caveat. Someone should be removed if they lose all of there games *with no moves played*. (Also, people should be removed from the infinity after losing one or at most two games with no moves played.)

I don't know how it is in other games, but in go the last around ten games take three to four extra months. And the worst is, when you look at the games they are decided and irrelevant.

An idea I like could be: When 25 games remain (from about 1200 in a tournament) they are scored by a computer program and finished immediately. Players are forced to accept the score. A free program like Fuego or Pachi could end the last 25 games in 25 minutes and save months of waiting to hundreds of players. I am sure many would volunteer some CPU cycles to do it.

And again, if you lose a tournament game by time you are removed from the tournament as if you hadn't played at all.

I definitely like the idea of auto removing inactive players from the next championship or infinity cycle. This will not speed up tournaments, but it will remove the clutter. I do not like erratic's idea of having an engine adjudicating go games. I have seen many positions that are supposed to be lost (go or other games), but can be won due to persistence, crafty play, and carelessness of the opponent.

I am not the slowest player, but I do play on the slow end of things. That is because I am a busy person. Sometimes I have no choice but to go for days without making moves, other days I make several moves. One of the reasons I like this site is because you do get as much time as you do. I say if you think the play is too slow, go to another site that has faster games.

The two things that do annoy me is when games go into negative time, and when the webmaster resets everyone's vacation. (vacation should be auto reset on Jan. 1 and that is it) These bugs do cause significant delays in tournaments. Luckily we have not seen negative time in awhile, but everyone did get vacation reset over the summer, so that can add 20 days to many tournaments right there. So it would be nice to see these bugs permanently fixed.

@ Ray Garrison: I also spend a lot of time in the first 100 moves or so because nine games (a whole group) are a lot of games and I cannot analyze them every day so I let some for the weekend, etc. I use all my time and really enjoy playing slow, but that does not require more than 6 months per game, say 9 at a maximum. My idea applies to say the worst 1% of the games (say the last 12 in 1200). Look at the games and tell me if they are worth playing for four extra months with hundreds of players waiting.

And anyway it's just an idea, not that I expect it to be implemented.

Since it's the early game that takes the most time we should have progressive time limits for some games. I think it would help if someone or something could adjudicate game that are running comparatively long or at least players should be encouraged via a message. For example this is the last game of the round please play quickly in courtesy to others. Ray why should quicker players find somewhere else to go?

Despite being a relatively fast player, I don't think we should be discriminating against slow players. Some people have busy lives, others may choose to invest a lot of thinking time on certain moves / games. The time control that LG offers is a relatively slow one, and players should be allowed to make the most of it if they wish.

Go 19 looks like a special case due to the length of the games, it's not a game I specialise in so I can't really add much to what has been said. Personally I don't notice the period where everyone is waiting for a couple of games to finish, as I always have other ongoing games, although I do sympathise with the specialist Go players here.

However I don't like the ideas of computer adjusication or encouraging players to play more quickly / resign. A game should be between the two players only, with no third party intervention. There may be promotion or relegation on the line. The slower player may have a 1% chance of winning. I think you just have to let these games play out, as frustrating as it is.

I agree with Martyn.

However, one possible adjustment could be made could be that as soon as the promotion/demotion status of each player in each section is determined, we start the next tournament.

I have noticed that some slower players are playing hundreds of games. It's no wonder they play slowly. Should there be some way of limiting number of games depending on the average time between moves?

Why not simply make tournament time controls faster/shorter, if the vast majority are finishing their games while a few players play on for months? That would seem to solve the problem without needing any new procedures/programming.

We HAVE a time limit here. And when you run out of time, you lose your games. I think that trying to punish slow players beyond that is a bad idea. Personally, I find 36 hours per move very suitable. Sure, sometimes I can play much faster (in most Einstein games), but sometimes I can't. In a difficult game I'll need time to think about my moves. Not 36 hours, by any means, but it's not every day I have the time to sit down for 15 or 30 mins to ponder over a game.

And the vacation days are necessary too, for obvious reasons. If you think the championships are slow, and I can agree that they sometimes are, just play other tournaments as well. Start RTs, or play in MCs. Really not a problem. So quicker players: Play RTs and MCs, don't leave LG. But don't try to alter LGs time rules either. That would ruin LG for many of us, and drive several players away from the site.

As for GO having the worst issues in this regards, that's partly because it's the slowest game. And partly because GO rules have certain weaknesses (particularly for non-experts), which can make them last more or less forever. Other games can have slow championships too, but GO is particularly slow due to these “flawed” rules.

I think that people who time out in a championship game should be unsubscribed from that championship, and sent a message telling them why, so that they can resubscribe if they still actually care.

I also like the idea of people who time out in games being forbidden from signing up for new games for a short period of time, so that they can't amass thousands of active games that they fail to play in.

I guess gerald would quit the site in protest ^^

Slow players should NOT be punished simply for being slow. Any game is between those two players (and an arbiter / tournament director, if appropriate) only. As frustrating as it might be for anyone else, it is not your business. Players play at irregular speed for many different reasons.

Having said that, we should be able to agree on improvements. Specifically, two suggestions made seem excellent (and I second them):

\* (@FatPhil) If a player times out in the championship, simply de-register them from the next championship. Send them a message informing them of this fact. They can always sign up again.

\* (@The_Shark_c) Start the next Championship as soon as the promotion / demotion status of all players is decided. If this means there are two or more championships running simultaneously, so what?

Small problem for the implementation of the above is that Richard appears to be otherwise distracted. Anybody know his status?

[Go19 is a special case as games can be VERY long, particularly between non-experts. If you're a Go19 specialist and are frustrated by slow play then you should play on a specialist Go19 site. Deciding rules specifically based on the experience of playing Go19 and then applying them to every other game is a poor idea.]

Go37x37 is a lot of fun, especially if you play against reactive players that play at least once a day. I accepted the invitation as a joke, but then I did not regret it (except of being bad at Go, meaning I should work more).

So the problem of slow players is not directly related to Go game, but to people who are a bit selfish and don't try to respect the moral contract you sign, when you sign in for a game.

Carroll: Where did you play Go 37x37?

I have left another site because they had stricter time limits. I don't think anyone would call me a slow player. (Possibly an occasionally slow player though.)

I also like the suggestions alain has listed again two posts above. But please no shorter time limits and no auto-ending of games.

Alain said: “Any game is between those two players (and an arbiter / tournament director, if appropriate) only. As frustrating as it might be for anyone else, it is not your business. Players play at irregular speed for many different reasons.”

This is just not true, it is my business if it restricts me from playing the games I love in a tournament setting. If you're playing a non-tournament game I couldn't care less how long you want to take but otherwise it affects me therefore it's my business.

I can appreciate that people come here to play somewhat slowly and adjudicating seems heavy handed, and progressive time limits might be difficult to implement but that doesn't mean we shouldn't be catering to the majority. Sure, chastise me for wanting to hurry people along but the most aggregiously slow players should also be chastened. I think a message to remind that last 10% of games that they are the last 10% would be sufficient to hurry a few games along.

Go19 is not the only game where this is a problem. Set up in DVONN takes forever and people playing out the end game in Amazons takes far too long.

@z: Here, at golem!

@z, I need an analyzer for Go37, cause my Goban is too small and I don't have enough stones…

@jugular: I do understand your frustration, but once a tournament starts, any game is between those players only, along with an arbiter or tournament director. Interfering, including complaining about their rate of play, is rightly frowned-upon in over-the-board tournaments, and similar should apply hrer. It is the responsibility, first and foremost, of the tournament director (in this case, Little Golem) to set up the tournament rules, in particular time controls, for a professional tournament which satisifies the requirements of all players. Testing ideas to improve future tournaments structure is a perfectly valid thing to do.

You are right that DVONN and Amazons can take a long time, good point. I still reckon Go19+ are at the extreme end of the spectrum.

So we need to come up with sensible suggestions we can all agree on as players and then see if LG will implement. So far I think these are ideas we can agree on:
\ De-register timeout players from next championship
\ Start next championship as soon as promotion / relegation is decided
\* Send message to players when 10% (and maybe 5% as well?) of games are outstanding

Anybody want to disagree, add something new (we can agree on!) or try to get Richard to implement?

These all look good to me. The only thing I'd add is regarding the first suggestion, when a Championship starts a player may be genuinely unaware of it. How about a warning (after 5 days perhaps) rather than just letting them time out? We're all here to play rather than win on time, aren't we?

I still think there will be issues with Go-19 due to the potential length of the game, but hopefully point 2 in particular will help to get the next Championship started sooner.

By the way, Dvonn and Amazons aren't too bad in my opinion, as there are only a limited number of moves before the board starts to vanish. The setup in Dvonn does take a while of course. For me, Shogi is the other Championship that takes a while to run.

> Send message to players…

There's already an e-mail message for each vacation day used. If players are ignoring them, I don't suppose any other warnings will be heeded.

It just appeared to me that this site has similarity to the world of Physics scientists:

The Nobelprize was given this year to a belgian and anglosaxian, for theoretically 'predicting' the existence of the 'God' particle giving everything weight. The Belgian clearly was earlier than the anglosaxian to produce his result, but who of both is 'Higgs'?

Anyway, I didn't want to distract of the discussion. I am not so hopeful towards the implementation, since I suggested the de-registration feature more than once in the past.

@jugular, in the normal McMahon go-tournament, you also have to wait until every single game of a round is finished before you can start the other round, even if it can sometimes be a little boring…

Otherwise I think simply starting the next tournament once promotions/relegations are set will speed the things up considerably (I know, I am one of those slower players, shame on me).

I've just caught up with this discussion.

I completely agree with the first two points of alain's suggestion. Deregister inactive players from the next championship, and start the next championship as soon as promotion and regulation have been decided. Many times the start of a new championship has been delayed by months because of only one or two games.

Because I have signed up for both Insanities, and always sign up to every Go Ahead, I have more games on the go than I am typically happy with, so I actually *like* the breaks between championships (which I do get, as I tend to be a fast player). If I desperately need a drip-feed of a game, there's always the infinity.

I think the first two suggestions are not controversial and all sides of the argument can agree on them. Problem is…where is Richard in all this? AFAIK, he's the only one who can develop LG code. Does anyone know him personally and can they suggest points 1,2 for near-future development?

> a belgian and anglosaxon
> The Belgian clearly was earlier

The Belgian was first for the field. The Anglo-Saxon was first for the particle.

Indeed, at GT and IYT the next round starts when all sections are decided but note that these servers take into account only points (no Berger coefficients) and only for the first place. This checking can be produced in a linear time. If we will try to determine medium places and to use Berger coefficients then to determine is a section decided or not in the worst case it becomes similar to satisfiability problem, so it seems to be NP-hard.

Well, apart from the fact that it's O(1), as every group has a constant size.

(I will note also that finding pre-images for SHA-2 is also an O(1) task, as is factoring 2048-bit RSA keys, so this isn't saying much.)

Every formula has a constant size. Nevertheless, the satisfiability problem is NP-hard.

Note also that a size of a group at LG is not restricted since the number of players added to a championship group due to their high ratings is not limited.

@Marsh Song: Yeah but even is StreetSoccer, I have never seen any group with more than 30 players.

Maybe we could start from the reduced problem i.e. tournaments with only one game or one player's games left, which seems much easier to be implemented ;)

@MarshSong: “N” is not constant in size. It's size grows monotonically with the size of its input N.

And I wouldn't think that it would be hard to find a case where promotion/demotion is undetectably knowable/unknowable. If a trivial algorithm works in all but pathelogical cases, then who gives a flying monkey bollock if the theoretical problem as a function of group size is NP-hard or not? For example if you just look at the extremal possible outcomes for each of the remaining players in isolation. Compare the old manual championship predictions that included the “if s/h/it wins all of his/her/its remaining games” column. Extend to “if s/h/it loses all his/her/its remaining games”, and you're a large part of the way there. The fact that the players with the most variability on outcome are the ones with the most remaining games left helps bound this problem.

@Ypercube: haven't the promotion/demotion rules been modified several times since the last megagroup was seen (I haven't seen one since my really really early days here)? I'm pretty sure that in the last few years I've been technically demoted from ch.1 and magically appeared back in ch.1, in at least one discipline, due to people dropping out. I presume that there's equivalent rebalancing to cope with people joining.

Personally, I don't see the benefit of having 2 championships being played in parallel. At least for the slower players. You make it sound like the problem is not enough games being played, when in reality the slow play is because there are too many games being played.

The problem is (as this only applies to go) the tournament takes over one year to finish and the players who play the last 4 or 5 months don't really use much time analyzing their games (because otherwise one of them would resign) but instead of playing fast they extend their time limits to the maximum expecting (maybe) that their opponent quits the server and they get a victory they don't deserve.

Playing another tournament is not a solution since nine simultaneous games take more time from work family, etc. than is reasonable (but we are game-junkies anyway). Playing 18 would get us both divorced and fired (or we could also play without really analyzing the game, but then, why on a turn based server? Isn't KGS a much better place for careless play?).

These players prove with their attitude that they don't really care much about the tournament. And for some reason I don't get they prefer to extend a lost game for six month using 100% of their time available plus twice vacation days (set to zero in winter and again in summer) rather than starting a new game they could actually win if they spent some time analyzing. If you have been analyzing the same game for months you reach a point where you can play it fast because you have already played out all the interesting variations long time ago.

So as I see it, this is more about people who care about the tournament vs. people who don't. And they probably wouldn't mind if they got their game scored and a new tournament with more games.

“Playing another tournament is not a solution” - with this I agree.

“they don’t really care much about the tournament” - this, alas, is mostly unproven. Sometimes the bad players don't see the terminal doom. I have no idea that there's an impending loss in games like 4-i-r, for example, and just mindlessly crank out “safe-ish” moves. We shouldn't penalise the less capable.
Deliberate time-wasters, yes, they should be castigated, but it's not always deliberate.

I go with Phil here. And in the end I think the challange is mostly about go19. I enjoy to breath between my first league CSs.

I'm with FatPhil on this one. I agree that this is a problem, at least for GO19-players, but erretic's solution seems worse than the problem to me.

I wish more of the players were more active in the forums, as I don't like the perceived them-and-us situation that arises occasionally, where “we” talk about “them” behind their backs. We should be politely and respectfully encouraging polite and respectful tournament behaviour, but that's hard when there isn't a common communication channel.

As always, the guys who kidnapped the go tournament for the last months are starting with the silly moves.

Eönwë's move 244 and Lieven Marchand's 231 are rude time losing moves with nothing to achieve. The four remaining games have been over for a long time.

Eönwë lost http://www.littlegolem.net/jsp/game/game.jsp?gid=1502045 by 5.5

Lieven Marchand lost http://www.littlegolem.net/jsp/game/game.jsp?gid=1502435 by 11.5 (+/- 1)

Huw Rees lost http://www.littlegolem.net/jsp/game/game.jsp?gid=1502535 by ¿30? Not even worth counting.

mellonsan lost http://www.littlegolem.net/jsp/game/game.jsp?gid=1502567 by 3.5-4.5 Ok. This looks like a close game, but there are not enough points in dispute for white to win. After 14 months, white never tried to actually win his game and convince himself that is it no longer possible. What do these guys spend their time doing if they don't care about trying to win their games?

I know it is rude to comment ongoing games but these games are long finished.

Game: http://www.littlegolem.net/jsp/game/game.jsp?gid=1502567&nmove=256

Before move 257 black had a solid 10 point lead (after some mistakes from white)

Ignoring white 256 is a huge mistake, since now white t2 change the bottom right from black to seki. Let's hope white plays it and black resigns his game out of embarrassment (even if the game becomes even, not lost).

Again, do these guys try to win their games? There hasn't been a decent game for months in the tournament.

Wow, dude.

Pretty poor form to kibitz on games in progress. They can read this, you know. One player to a hand.

These are amateur players. They're going to make mistakes. In correspondence games, it's not always easy to remember board positions.

Not everybody is as good as you are. Try not to let it upset you, and you'll be happier for it.

I agree with pfafulous! kibitzing is good for teaching, but NEVER should occur during a tournament game.

The lesson of Go should be …..patience! The game WILL end eventually. Be patient!

One solution on Boite a Jeux to the issue of slow play is to disallow entry into new games if it is your move in 25% of more of your current active games. In addition, everyone has a speed rating, and premium members of that site can screen slower players from new games. These solutions are straightforward on BaJ because of its emphasis on individual, rated games rather than tournaments.

Nevertheless, I see that an issue on LG is that of players with hundreds (or thousands!) of ongoing active games, and perhaps it would work to restrict entry into new games, including tournaments, with at least 25% of existing games awaIting your turn.

I agree with Ray Garrison that patience is a virtue, but it's bad for LG if tournament players are inactive for month after month while waiting for a few games to grind to an inevitable conclusion.
Luckily, the issue appears to be largely restricted to Go.

I would not comment ongoing games normally. But these games started in December 2012 and 300 players are waiting for them to finish so a new tournament can start.

@erratic: Go Hahn System championship is almost ended. What about play there too or start to play in infinity tournament.

“disallow entry into new games if it is your move in 25% of more of your current active games” – this gets my strong support.

Waived (or with a different threshold) for paid up members, perhaps; I have no problem with that. (We freeloaders get things way too easy as it is.)

Hi,

*de-register timeout players from championships*
Sounds good to me. So what could be the condition for the deregistration?
My proposal: Timeout in any game of a particular championship triggers deregistration of that championship. (So any timeout in the Go19 championship leads to deregistration from the Go19 championship, but not from other championships.)

*start a new tournament as soon as promotion/demotion is clear*
This sounds partially good. I find it hard to implement such a feature, because the slowest players tend to have many unfinished games. This means that they can still gain many points and can end up anywhere in the ranking except maybe the top and bottom position. An important exception are, of course, the lowest levels of a championship, where only promotion needs to be clarified. Here, I agree that the tournament could be ended, e.g. by forced ties, when the first ranks in the lowest championship level are clear.

I'd like to make two new proposals regarding vacation days.
*more frequent replenishment of vacation days*
Instead of giving every player a certain number of vacation days at the beginning of each year (and also maybe during the year due to some technical issues), I propose to give a certain number of vacation days at the beginning of each month(e.g. 2 for puny earthlings like me, 3 for LG members). At the same time, limit the number of vacation days to 20 (or 30 for members).

*taking vacation only affects “due” games*
Instead of adding 24 hours to every game of a player who is taking one day of vacation, only add the 24 hours to games that have less than _N_ hours left (with N=24,36,48 or whatever threshold, currently we have N=infinity). The evident benefit is that you will have your 20 (or 30) vacation days, but when a player has used up all his vacation days in a row, there are no game

…
there are no games with more than _N_ plus 24 hours left.

I have still one proposal.
*hurry up slow players*
Make the award for a played move in a game dependent on the “type” of tournament. Currently, LG has the “+36 hours”-rule for all tournaments (championship, monthly cup, rated game, unrated game). I'd propose to introduce a “hurry up” flag for championships. When _N_ percent of all games are finished (e.g. N=90), change the championship into a “quick mode”, where the award for a played move is only +24 hours. This could speed up the end of a championship.

I also like the earn-vacation-days-monthly idea. In particular if you don't earn them if you've timed out of any games that month. The moribund must be put out of their, or our, misery more quickly.

Different disciplines could have different time constraints, and thos time constraints need not be constant. Real world chess is the classic example.

I have a somewhat related idea about vacations.

Everyone starts with %I hours of vacation. Each move is expected to be made within %E hours when it's your turn. If you make a speedy move in %s hours (%s <= %E), you accrue %R * (%E - %s) hours of vacation; otherwise, you lose %P * (%s - %E) hours. Vacation is capped at %M hours, and if you use it up, all games awaiting your move time out.

Suggested parameters:
%I = 36 (initial vacation hours)
%E = 36 (expected delay per move)
%R = 0.25 (reward coefficient)
%P = 1.5 (penalty coefficient)
%M = 240 (maximum vacation hours)

This proposal not only imposes a slow-move penalty but also offers an incentive to fast players.

@z
I first felt comfortable with the reward for speedy moves, but quickly realized that it can be misused for deliberately delaying tournaments.

I dislike the concept of penalties, because:
The penalty part (with the current parameters) means that when, for example, you
- have 120 hours left in a game (5 days)
- can not play for 96 hours (4 days) and then make your move
your remaining amount of hours in that game drop from 24 to -6 due to that penalty. So you play in time and still will lose one day of vacation.

On the other hand it could be reasonable just to *limit the amount of games that can be played simultaneously* on Little Golem. I am mentioning this, because I currently play Go9 against an opponent who is awfully slow (and the game is already decided, in my opinion). He is making a move every 3 to 4 days in average, and is hence taking vacation days, of course. Why is he playing so slowly? He currently has these games in progess:
273 Einstein würfelt nicht,
56 Street Soccer,
45 Shogi,
21 Lines of action,
18 Go
and some others, so I understand that my opponent is quite busy.

By limiting, I do not mean simply limit the number of games. Introducing a weighted system (e.g. 10 points for Go9, 14 points for Go13, 20 points for Go19, and so on) together with a limit of the total points per player (e.g. 1000 points) could prevent an overload for players (if they do not resort to multiple accounts). This is roughly the policy: *An average player should not spend more than _N_ hours daily playing on Little Golem*

Stopping poeple from signing up or being signed up for new events unless they've played enough of their moves, as mentioned above, helps keep some players like that more under control. 400+ games is pretty wild. I'm uncomfortable with more than about a third of that.

+1 on limiting the total amount of active games.

On second thought, my earlier proposal is overly complicated and may be less effective at speeding up tournaments.

The point is to limit signup when players have many games active when it is their move. There may be one or two players whose life is LG, and who can keep a thousand games up to date…..

A long time ago I used to have 700+ ongoing games :)

I like the rules they are, especially their simplicity. I don't like non-local rules, e.g. making time depend on the status of other games. (Imagine your clock shows 12 hours, you go to sleep and in the morning you lost just because some other games are finished now!?) I don't like limitations like a maximum number of ongoing games. Not adding vacation days to all games is against the idea of vacation. Playing more than one championship at the same time is against the idea of championship, against the idea of having a single champion.

I don't like any kind of special treatment for people that follow the rules.

We could, instead, change the default and automatically de-register all players on tournament start. With other words: A subscription would be valid for a single tournament only. Seems to be the solution that is easiest to implement and could be agreed on.

Instead of a pool of 240 hours, I like byi yomi time (AKA grace time). Give everyone 36 hours that gets refreshed whenever they make a move, but also a pool for each game of 204 hours (240 - 36) that doesn't get refreshed in a game. Change the numbers, increasing the 36 and perhaps decreasing the 204 if you want to accommodate players who have more frequently irregular schedules. But making the whole pool of 240 hour refreshable just encourages procrastination followed by stress, I think. Vacation would still handle real absences.

I don't see how restricting the number of games will necessarily make slow players move more quickly. It appears some players will always make their moves at the last possible time.

@Avins I agree a shorter time makes sense but there does not appear much support in the LG community.

So two other suggestions:
1) Have both fast and regular time tournaments.
2) Make vacation time opt in. You have to elect to go on vacation and if you don't the game is forfeited.

Opt in for vacation time is the standard on many other sites. That it is automatic here is one of the main reasons for me of playing here and not elsewhere.

is it possible to have a “real time area” on this side? So that two player can play in hours a 50 Point game?

May be it is a variant for fast player

There is a suggestion to add a vacation day every 18 days (up to a limit of 20 vacation days, newly registered players start with 10 days),

365 days / 18 days = 20,27 vacation days.
(Members get a vacation day added every 12 days)

It would not solve everything for slow players, but a newly registered player in e.g. November could not use 40 vacation days in a row (only about 10).

**
The other problem is in games like Go (or e.g. Havannah), they consist of many moves, which is the reason they can take so long.

This is, as far as I know, one of the big discussions of why tournaments can drag on, and why games last longer than they're expected to. The other is http://littlegolem.net/jsp/forum/topic2.jsp?forum=1&topic=693.

If other people know of other such discussions that are worth preserving, link them here. I mention this both to bump the thread, and because it's come up again: http://littlegolem.net/jsp/forum/topic2.jsp?forum=55&topic=315.