Daniel Faken | 3 Oct 20:10 2005

Problems with latest updates

Hello,

  Am I the only one for whom the latest updates completely mess up
Croquet?
  I get a huge horizontally-floating rectangle covered with a
texture-array of Croquet-developers' faces, the avatar's cursor is huge,
centered, and blocks a lot of other things.  There appears to be some kind
of 2D rendering (of a smiley?) moving around in the ULH corner, etc. etc.

I also get errors when loading 3DSMax files (presumably for the 'dungeon'
world), in the standard TeapotMorph.

This is with updates through 0280.

I previously had updates through 0274, and had no such problems.

thanks!
Daniel

Ed Boyce | 3 Oct 22:44 2005

Re: Problems with latest updates

Daniel,

  The update crashes me before I can really do much when I apply it.
(I.e., the system crashes so fast I can,t even save the changes.

  Note that I'm trying to update a ChemStereo Croquet image on Linux. 
This hasn't caused me trouble before, but it may be a commonality
between our installations if no one else is seeing the trouble.

    -- Ed

Daniel Faken said:
> Hello,
>
>   Am I the only one for whom the latest updates completely mess up
> Croquet?
>   I get a huge horizontally-floating rectangle covered with a
> texture-array of Croquet-developers' faces, the avatar's cursor is huge,
> centered, and blocks a lot of other things.  There appears to be some kind
> of 2D rendering (of a smiley?) moving around in the ULH corner, etc. etc.
>
> I also get errors when loading 3DSMax files (presumably for the 'dungeon'
> world), in the standard TeapotMorph.
>
> This is with updates through 0280.
>
> I previously had updates through 0274, and had no such problems.
>
> thanks!
> Daniel
(Continue reading)

Howard Stearns | 3 Oct 23:53 2005
Picon

Re: Problems with latest updates

I'm not having any problems with the current updates. I'm using a Mac.

I do often get pretty bizarre errors if I run something that put things 
in the data directory before making a code change.  Furthermore, 
there's at least one bug in Squeak's idea of what directories are 
present.  So.... When I get a truly bizarre and unexplained error, the 
first thing I do (after determining in the debugger that there's no 
good explanation for the error) is quit Squeak, delete the data 
directory, restart Squeak, and try again.

Howard Stearns
Croquet Lead Developer, http://croquetproject.org
DoIT Academic Technology, http://www.wisc.edu/academictech
University of Wisconsin-Madison, 1301 University Avenue, Madison, WI  
53715
  +1-608-262-3724

On Oct 3, 2005, at 3:44 PM, Ed Boyce wrote:

> Daniel,
>
>   The update crashes me before I can really do much when I apply it.
> (I.e., the system crashes so fast I can,t even save the changes.
>
>   Note that I'm trying to update a ChemStereo Croquet image on Linux.
> This hasn't caused me trouble before, but it may be a commonality
> between our installations if no one else is seeing the trouble.
>
>     -- Ed
>
(Continue reading)

Daniel Faken | 4 Oct 03:36 2005

Re:Problems with latest updates

Well, it looks like its somehow my fault (coming from difference btw. my image & the standard one).

Sorry!

thanks,
Daniel

David Faught | 4 Oct 22:33 2005
Picon

TeaTime analogy

Refering to Howard Stearns' blog entry at http://www.wetmachine.com/itf/item/334 which describes a
very nice analogy for TeaTime, here is another analogy that might provide some fodder for further
discussion, or maybe not.

In a different but related vein to what Howard describes, my teenage daughter spends lots of her computer
time using AIM, with several independent but concurrent conversations with her friends.  There is no
shared base state here, other than direct knowledge of the people on the other ends of the conversations,
so the emphasis is on the content of the conversations.  Each conversation is truly independent of the
others, except where the conversants choose to share information across conversations.

From my daughter's point of view this is a one-to-many relationship, except that each of the many is
separate and the conversations distinct.  There usually is, in fact, a web of conversations as each of the
conversants may have their own set of other friends connected.

This is a high overhead situation, and not particularly adept at sharing one idea with multiple
conversants, although much better than the telephone conversations of my teenage years.  It is, on the
other hand, very good at isolating multiple concurrent conversations and thereby dealing with each
conversant's individual characteristics and, at least somewhat, security issues.

So, I guess one question is, how does TeaTime-based messaging deal with multiple conversants with varying requirements?

orion elenzil | 4 Oct 23:17 2005
Picon

Re: TeaTime analogy

David Faught dave_faught-at-yahoo.com |squeak devlists| wrote:

 > So, I guess one question is,
 > how does TeaTime-based messaging deal with multiple conversants with varying requirements?

one good example of this is suppose i have some data
(which i maintain does occasionally exist, pax david smith! :)
such as an image or a file,
and i want to share it with my collaborateurs.

now,
some of my collaborateurs already have the file in some cache, and some don't.
it's a waste of my bandwidth and more egregiously a waste
of people's who already have the content if i send it out to everybody.

how will teatime arrange for this ?

Orion

PS
many folks already know,
but i should add that i'm more or less dropping out of the Croquet community,
as i've taken a 13-month position at the south pole of the earth.
I'll still be following the discussions,
but i doubt i'll have time to do any actual development.
see my blog http://orionelenzil.blogspot.com

David Faught | 5 Oct 04:09 2005
Picon

re:TeaTime analogy

Okay, this isn't a strong analogy because the point of TeaTime is shared state between participants, which
is not so with AIM.  But there may be differences between the participants and even potential differences
between the desired state of each participant depending on security levels or access, for example.

One of the major problems has to be how to catch up late joiners, since the desired shared state is not
necessarily just a straight function of time for each object.  I recall seeing some methods attempting to
deal with this in Croquet Jasmine by repeated quick steps with no wait time between, but I think this didn't
really work because the late joiner could just seem to loop indefinitely depending on how far it was behind.

I'm just thinking out loud here, but I'm also curious ...

David Faught | 5 Oct 14:34 2005
Picon

re:TeaTime analogy

I have received more off-list response to this than on-list.  I'm guessing that one reason is that no one
(including me) wants to distract the Croquet Architects and the closed development team, who might
actually know some of the answers.  So, I'll just leave off on this topic until we, the public Croquet
community, actually find out what the answers are.

Slowly turning blue (not really !-)
Dave

nick hemsley | 5 Oct 15:36 2005
Picon

Re: TeaTime analogy

Oops, must remember to reply-all.

I dont quite understand your comment about catching up late joiners, are you saying that objects may depend upon recieving messages at a certain time, relative to the last message? I assume in this case, any calls to the system time can be wrapped & return the virtual time of the message.

Reading Howard Stearns blog article (a while ago now) got me thinking about how the actual teatime session is replicated,  specifically: what events need to be replicated. I have looked into it more & it seems (very preliminary assessment by me ) that every subclass of the TObject is possibly involved in the replication process. Does anyone know if this is correct?

Howard's blog seemed to indicate that it is IO (mouse, keyboard, HDD perhaps) events that need to be replicated & then this sets off the required chain of events that results in the state being replicated. Although, it just occured to me that this does not allow any errors to be caught. So I guess i am back at my assessment, arrived at by looking at the architecture, that every subclass of the TObject takes care of replication, or at least registers itself in the teatime architecture, or whatever.

Couple of ideas as well, suppose that a session of users are working collaboratively on a complex problem (i.e. rendering a film & editing realtime in a croquet space), if each party is involved in rendering the scene, then the rendering algorithm can only proceed as the fastest (or slowest?) node. Does anyone see the need for distributed computations in a shared teaparty? This also extends to the situation where  a persistent session may span thousands of users & GB of object state, but I think this has been talked about by howard stearn somewhere & also touched on in the discussion about voronai graphs & sharing state.

Cheers.

On 10/5/05, David Faught <dave_faught <at> yahoo.com> wrote:
I have received more off-list response to this than on-list.  I'm guessing that one reason is that no one (including me) wants to distract the Croquet Architects and the closed development team, who might actually know some of the answers.  So, I'll just leave off on this topic until we, the public Croquet community, actually find out what the answers are.

Slowly turning blue (not really !-)
Dave



Ralph Johnson | 5 Oct 16:08 2005
Picon

Re: TeaTime analogy


>Date: Wed, 05 Oct 2005 21:36:02 +0800
>From: nick hemsley <lists.nick.hems <at> gmail.com>
>
>Couple of ideas as well, suppose that a session of users are working
>collaboratively on a complex problem (i.e. rendering a film & editing
>realtime in a croquet space), if each party is involved in rendering the
>scene, then the rendering algorithm can only proceed as the fastest (or
>slowest?) node. Does anyone see the need for distributed computations in a
>shared teaparty? This also extends to the situation where a persistent
>session may span thousands of users & GB of object state, but I think this
>has been talked about by howard stearn somewhere & also touched on in the
>discussion about voronai graphs & sharing state.
I think that if one of the nodes is not fast enough, it gets bad
results, but it does not effect the other nodes.  People will try
to keep their scenes simple so that those with slower computers
will not complain, but I think the only effect of the slow computers
on the fast ones is social, not technical.

David Reed is going to talk about Teatime at OOPSLA the week after
next.  If any of you are going then you should be sure to talk to
him about it.  Maybe ask questions after the talk.  OOPSLA is in
San Diego, so if you live in the area, you might register for the
day of his talk.  There is going to be a Croquet workshop, too.

-Ralph Johnson


Gmane