Large Memory Java Applications

[Owen Densmore]

I started a discussion on large memory java applications on Java  
Lobby.  Basically how to approach graph problems with 250 million  
nodes.  The response was surprisingly good!
   http://www.javalobby.org/java/forums/t72726.html

It seems there are two approaches:
- Hardware: build a large server or a cluster and use interesting  
distributed solutions that have become available for this
- Software: use interesting streaming|piping solutions, or possibly  
DB-like disk solutions that can be tuned for your application.

Has anyone here tackled truly large problems like this successfully?

     -- Owen

Owen Densmore
http://backspaces.net - http://redfish.com - http://friam.org

[Dan Kunkle]

For my thesis work, I'm looking at doing very large graph search to
find optimal solutions to Rubik's cube. As a component of our
approach, we're planning to do a full breadth-first search of a graph
with over 1 trillion nodes.

To do this, we're using the combined disk space of a cluster of
machines. Of course, disks are slow if you use random access, so you
have to convert all of your algorithms to using streaming access only
(depending on the algorithm, this can be very tricky). In fact, with
today highly non-uniform memory hierarchies (L1 cache -> L2 cache ->
... -> main memory), even using RAM by way of random access can make
large computations infeasible. So, these same streaming-access-only
algorithms can also provide a massive speedup when disk isn't used.
Further, if you use streaming access only with disk, the aggregate
disk bandwidth can equal that of the bandwidth to main memory. So, if
you have a cluster of 100 machines with 300 GB of disk each, this
means your cluster works like a single machine with 30 TB of RAM.

Depending on whether you have an explicit graph (the whole thing
stored in memory) or an implicit graph (defined by a function
providing the neighbor nodes of any given node), the techniques may
differ. I'm more familiar with implicit graph search, but it sounds
like you may be working with explicit graphs.

Here is a reference from each camp dealing with very large searches in
external memory:

AI community -- implicit graph searches:
Korf and Schultze, Large-Scale Parallel Breadth-First Search
http://www.cs.ualberta.ca/~bulitko/F05/CMPUT651/papers/pbfs.pdf

Explicit graph manipulation:
STXXL, (Standard Template Library for Extra Large Data Sets (C++ library))
http://stxxl.sourceforge.net/

I don't have any Java specific advice, but the key of using only
streaming access with large data structures is language independent.

-Dan

--
[ http://www.ccs.neu.edu/home/kunkle/ ]

On 5/26/06, Owen Densmore <owen at backspaces.net> wrote:

> I started a discussion on large memory java applications on Java
> Lobby.  Basically how to approach graph problems with 250 million
> nodes.  The response was surprisingly good!
>    http://www.javalobby.org/java/forums/t72726.html
>
> It seems there are two approaches:
> - Hardware: build a large server or a cluster and use interesting
> distributed solutions that have become available for this
> - Software: use interesting streaming|piping solutions, or possibly
> DB-like disk solutions that can be tuned for your application.
>
> Has anyone here tackled truly large problems like this successfully?
>
>      -- Owen
>
> Owen Densmore
> http://backspaces.net - http://redfish.com - http://friam.org
>
>
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org
>

... [show rest of quote]

[Marcus G. Daniels-2]

Hi Dan, that seems like a clever approach..  One remark on this bit:
> In fact, with
> today highly non-uniform memory hierarchies (L1 cache -> L2 cache ->
> ... -> main memory), even using RAM by way of random access can make
> large computations infeasible.
By Little's law (http://en.wikipedia.org/wiki/Little's_law), concurrency
= bandwidth * latency.
If it's possible to parallelize the search, then the latency of the
memory system can be tolerated, as there are more threads of execution
that can do useful work.  Chances are they won't all be blocked on
memory access.  For example, with a distributed shared memory package on
an Infiniband cluster (e.g. Intel's Cluster OpenMP*), or a highly
multithreaded CPU architecture (UltraSparc T1), I think there's still a
practical hope not to give up on random access algorithms.

Marcus

[Jim Rutt]

Indeed.   A grid is not a grid is not a grid.  A loosely coupled "roomful
of linux boxes" running on a couple legs of gigabit ethernet has far
different attributes than  does a closely coupled system like the IBM SP2
or the the Cray XD1.

The one and only "really big cluster" thing I did was back in the late 90s
we bought one of the first really big Sun T10000 configurations, put 100
gig of memory in it, and loaded the most heavily used part of our Westlaw
legal database.  $2 million of Sun stuff replaced $50million of Plug
Compatible VM mainframes.  That project, though was done using C and a
homebrew in memory knockoff of adabase.

=jim




At 10:16 AM 5/26/2006, you wrote:

>Hi Dan, that seems like a clever approach..  One remark on this bit:
> > In fact, with
> > today highly non-uniform memory hierarchies (L1 cache -> L2 cache ->
> > ... -> main memory), even using RAM by way of random access can make
> > large computations infeasible.
>By Little's law (http://en.wikipedia.org/wiki/Little's_law), concurrency
>= bandwidth * latency.
>If it's possible to parallelize the search, then the latency of the
>memory system can be tolerated, as there are more threads of execution
>that can do useful work.  Chances are they won't all be blocked on
>memory access.  For example, with a distributed shared memory package on
>an Infiniband cluster (e.g. Intel's Cluster OpenMP*), or a highly
>multithreaded CPU architecture (UltraSparc T1), I think there's still a
>practical hope not to give up on random access algorithms.
>
>Marcus
>
>
>============================================================
>FRIAM Applied Complexity Group listserv
>Meets Fridays 9a-11:30 at cafe at St. John's College
>lectures, archives, unsubscribe, maps at http://www.friam.org

... [show rest of quote]

===================================
Jim Rutt
voice:  505-989-1115

[Marcus G. Daniels-2]

Jim Rutt wrote:
> $2 million of Sun stuff replaced $50million of Plug
> Compatible VM mainframes.  That project, though was done using C and a homebrew in memory knockoff of adabase.
>  
Ah, right, then there's the question of what kind of database that could
store all of the data and access it fast enough (either for streaming or
for parallel traversal).

Someone in our group went to the recent MySQL conference and heard
mention of their `Falcon' backend (that I guess they started because
Oracle bought up InnoDB) that's in development.   Below is a link to  
some notes on the talk, sounds intriguing to me..

http://mike.kruckenberg.com/archives/2006/04/jim_starkey_int.html

[Owen Densmore]

Hi Dan, good to hear from you.

So, given your understanding of RedFish .. would you recommend some  
sort of "super computing" stunt for us?  Either a cluster or a  
biggish multiprocessor 64-bit linux/solaris box?

We've been considering such a step, or hoping that there becomes  
available some sort of "compute farm" much like the Blender render  
farm we're using for great visualization work.

     -- Owen

Owen Densmore
http://backspaces.net - http://redfish.com - http://friam.org


On May 26, 2006, at 9:43 AM, Dan Kunkle wrote:

> For my thesis work, I'm looking at doing very large graph search to
> find optimal solutions to Rubik's cube. As a component of our
> approach, we're planning to do a full breadth-first search of a graph
> with over 1 trillion nodes.
>
> To do this, we're using the combined disk space of a cluster of
> machines. Of course, disks are slow if you use random access, so you
> have to convert all of your algorithms to using streaming access only
> (depending on the algorithm, this can be very tricky). In fact, with
> today highly non-uniform memory hierarchies (L1 cache -> L2 cache ->
> ... -> main memory), even using RAM by way of random access can make
> large computations infeasible. So, these same streaming-access-only
> algorithms can also provide a massive speedup when disk isn't used.
> Further, if you use streaming access only with disk, the aggregate
> disk bandwidth can equal that of the bandwidth to main memory. So, if
> you have a cluster of 100 machines with 300 GB of disk each, this
> means your cluster works like a single machine with 30 TB of RAM.
>
> Depending on whether you have an explicit graph (the whole thing
> stored in memory) or an implicit graph (defined by a function
> providing the neighbor nodes of any given node), the techniques may
> differ. I'm more familiar with implicit graph search, but it sounds
> like you may be working with explicit graphs.
>
> Here is a reference from each camp dealing with very large searches in
> external memory:
>
> AI community -- implicit graph searches:
> Korf and Schultze, Large-Scale Parallel Breadth-First Search
> http://www.cs.ualberta.ca/~bulitko/F05/CMPUT651/papers/pbfs.pdf
>
> Explicit graph manipulation:
> STXXL, (Standard Template Library for Extra Large Data Sets (C++  
> library))
> http://stxxl.sourceforge.net/
>
> I don't have any Java specific advice, but the key of using only
> streaming access with large data structures is language independent.
>
> -Dan
>
> --
> [ http://www.ccs.neu.edu/home/kunkle/ ]
>
> On 5/26/06, Owen Densmore <owen at backspaces.net> wrote:
>> I started a discussion on large memory java applications on Java
>> Lobby.  Basically how to approach graph problems with 250 million
>> nodes.  The response was surprisingly good!
>>    http://www.javalobby.org/java/forums/t72726.html
>>
>> It seems there are two approaches:
>> - Hardware: build a large server or a cluster and use interesting
>> distributed solutions that have become available for this
>> - Software: use interesting streaming|piping solutions, or possibly
>> DB-like disk solutions that can be tuned for your application.
>>
>> Has anyone here tackled truly large problems like this successfully?
>>
>>      -- Owen
>>
>> Owen Densmore
>> http://backspaces.net - http://redfish.com - http://friam.org
>>
>>
>>
>> ============================================================
>> FRIAM Applied Complexity Group listserv
>> Meets Fridays 9a-11:30 at cafe at St. John's College
>> lectures, archives, unsubscribe, maps at http://www.friam.org
>>
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org

... [show rest of quote]

[Owen Densmore]

Having worked at Sun, where an E10,000 or three were available for  
our use in the labs (64 processor, huge memory), I remember  
rebuilding a huge shell script to be able to run in parallel a power  
law exploration:
   http://www.backspaces.net/sun/PLaw/

The job went from over night on a really fast laptop, to 15 minutes.  
Looking at the multi-processor perfmeter was like looking on a bunch  
of boxes catch fire!

     -- Owen

Owen Densmore
http://backspaces.net - http://redfish.com - http://friam.org


On May 26, 2006, at 11:23 AM, Marcus G. Daniels wrote:

> Jim Rutt wrote:
>> $2 million of Sun stuff replaced $50million of Plug
>> Compatible VM mainframes.  That project, though was done using C  
>> and a homebrew in memory knockoff of adabase.
>>
> Ah, right, then there's the question of what kind of database that  
> could
> store all of the data and access it fast enough (either for  
> streaming or
> for parallel traversal).
>
> Someone in our group went to the recent MySQL conference and heard
> mention of their `Falcon' backend (that I guess they started because
> Oracle bought up InnoDB) that's in development.   Below is a link to
> some notes on the talk, sounds intriguing to me..
>
> http://mike.kruckenberg.com/archives/2006/04/jim_starkey_int.html
>
>
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org

... [show rest of quote]

[Tim Densmore]

Speaking of blender, this is very neat and produced (IIRC) entirely with with
blender.  The story is short and very surreal, but I was quite impressed.

http://orange.blender.org/


On Friday 26 May 2006 20:21, Owen Densmore wrote:
> Blender

[Stephen Guerin]

We have the DVD of Elephants Dream in the office and will play it before
Carlos Gershenson's talk on Wednesday.

Interestingly, all .blend files used to create the movie are included on the
DVD and downloadable from the web.

-Steve

> -----Original Message-----
> From: Tim Densmore [mailto:tim at backspaces.net]
> Sent: Friday, May 26, 2006 9:15 PM
> To: The Friday Morning Applied Complexity Coffee Group
> Subject: Re: [FRIAM] Large Memory Java Applications - OFF TOPIC
>
> Speaking of blender, this is very neat and produced (IIRC)
> entirely with with blender.  The story is short and very
> surreal, but I was quite impressed.
>
> http://orange.blender.org/
>
>
> On Friday 26 May 2006 20:21, Owen Densmore wrote:
> > Blender
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org
>
>

... [show rest of quote]

[Tim Densmore]

http://www.tllts.org/dl.php

That's a link to a recent interview with Bassam Kurdali, the director of ED.  
It happens to be on my favorite leisure time podcast, The Linux Link Tech
Show.  The sound quality is often very poor, and the show tends to be very
freeform, but I've been listening to their show for nearly 2 years - they
must be doing something right!  Beware of the occasional right-wing and/or
Christian fundamentalist insert by Linc or Pat.

On Friday 26 May 2006 21:29, Stephen Guerin wrote:

> We have the DVD of Elephants Dream in the office and will play it before
> Carlos Gershenson's talk on Wednesday.
>
> Interestingly, all .blend files used to create the movie are included on
> the DVD and downloadable from the web.
>
> -Steve
>
> > -----Original Message-----
> > From: Tim Densmore [mailto:tim at backspaces.net]
> > Sent: Friday, May 26, 2006 9:15 PM
> > To: The Friday Morning Applied Complexity Coffee Group
> > Subject: Re: [FRIAM] Large Memory Java Applications - OFF TOPIC
> >
> > Speaking of blender, this is very neat and produced (IIRC)
> > entirely with with blender.  The story is short and very
> > surreal, but I was quite impressed.
> >
> > http://orange.blender.org/
> >
> > On Friday 26 May 2006 20:21, Owen Densmore wrote:
> > > Blender
> >
> > ============================================================
> > FRIAM Applied Complexity Group listserv
> > Meets Fridays 9a-11:30 at cafe at St. John's College
> > lectures, archives, unsubscribe, maps at http://www.friam.org
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org

... [show rest of quote]

[Tim Densmore]

Oh, there's also a fair amount of "inappropriate language."

On Friday 26 May 2006 21:51, Tim Densmore wrote:

> http://www.tllts.org/dl.php
>
> That's a link to a recent interview with Bassam Kurdali, the director of
> ED. It happens to be on my favorite leisure time podcast, The Linux Link
> Tech Show.  The sound quality is often very poor, and the show tends to be
> very freeform, but I've been listening to their show for nearly 2 years -
> they must be doing something right!  Beware of the occasional right-wing
> and/or Christian fundamentalist insert by Linc or Pat.
>
> On Friday 26 May 2006 21:29, Stephen Guerin wrote:
> > We have the DVD of Elephants Dream in the office and will play it before
> > Carlos Gershenson's talk on Wednesday.
> >
> > Interestingly, all .blend files used to create the movie are included on
> > the DVD and downloadable from the web.
> >
> > -Steve
> >
> > > -----Original Message-----
> > > From: Tim Densmore [mailto:tim at backspaces.net]
> > > Sent: Friday, May 26, 2006 9:15 PM
> > > To: The Friday Morning Applied Complexity Coffee Group
> > > Subject: Re: [FRIAM] Large Memory Java Applications - OFF TOPIC
> > >
> > > Speaking of blender, this is very neat and produced (IIRC)
> > > entirely with with blender.  The story is short and very
> > > surreal, but I was quite impressed.
> > >
> > > http://orange.blender.org/
> > >
> > > On Friday 26 May 2006 20:21, Owen Densmore wrote:
> > > > Blender
> > >
> > > ============================================================
> > > FRIAM Applied Complexity Group listserv
> > > Meets Fridays 9a-11:30 at cafe at St. John's College
> > > lectures, archives, unsubscribe, maps at http://www.friam.org
> >
> > ============================================================
> > FRIAM Applied Complexity Group listserv
> > Meets Fridays 9a-11:30 at cafe at St. John's College
> > lectures, archives, unsubscribe, maps at http://www.friam.org
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org

... [show rest of quote]

[Steve Smith]

> Ah, right, then there's the question of what kind of database that
> could
> store all of the data and access it fast enough (either for streaming
> or
> for parallel traversal).
During my year at LBL last year, I worked with the principles of this
system in the Scientific Data Management group:

http://sdm.lbl.gov/fastbit/

We've been talking about it's application to UltraScale graphs since I
got there...

One motivation I have is to precompute various "local" graph theoretic
measures to use later in heuristics for various larger-scale searches,
measure and layout.  A blindingly obvious application of this
technology is "graph conditioning", or extracting a subgraph based on a
set of conditions.

One good "trick" we discovered was to derive ubiquitous properties of a
raw data set, index the result in a way that compresses well (the
resulting index uses a tiny fraction of the space required for the
indexed quantitiy) and then discard the derived quantity...  When a
search returns a set of "candidates" for matching, the (often tiny
subset of) the results can then be used to recalculate the derived
property.  It leads to  an interesting multi-step tradeoff between
computation and storage.

FastBit uses a (patent pending) Word Aligned Hybrid Compression of Bit
Indices.  We discovered all kinds of nice (read potentially useful)
properties of the compressed indices themselves, including providing a
rough entropy measure of the data being indexed.

FastBit shines in high-dimensional range-queries... so like, give me
the call graph of all the phone calls in England originated between
9:00 PM and 9:23PM GMT which lasted for more than 10 seconds and less
than 2 minutes with differing area codes but excluding those
originating from 555 or terminating in a number above 799 ...

I'm hoping it can be made to be meaningful in the context of
ultra-scale graphs as well... I'm thinking it will require another
level of indirection...

- Steve

[Marcus G. Daniels-2]

Steve Smith wrote:
> FastBit shines in high-dimensional range-queries... so like, give me
> the call graph of all the phone calls in England originated between
> 9:00 PM and 9:23PM GMT which lasted for more than 10 seconds and less
> than 2 minutes with differing area codes but excluding those
> originating from 555 or terminating in a number above 799 ...
>  
Is FastBit source code freely redistributable or is there a commercial
implementation?   I can't find mention of either on the web page.

[Steve Smith]

On May 27, 2006, at 4:33 PM, Marcus G. Daniels wrote:

> Steve Smith wrote:
>> FastBit shines in high-dimensional range-queries... so like, give me
>> the call graph of all the phone calls in England originated between
>> 9:00 PM and 9:23PM GMT which lasted for more than 10 seconds and less
>> than 2 minutes with differing area codes but excluding those
>> originating from 555 or terminating in a number above 799 ...
>>
> Is FastBit source code freely redistributable or is there a commercial
> implementation?   I can't find mention of either on the web page.
>

Right now (as I understand it) the answer (as you may antiicpate) is  
neither.

This does not necessarily mean that someone cannot use their binaries  
in a non-commercial system...  or ultimately license their software  
if you want to use it commercially.... or learn something completely  
new from their approach and work it from a different angle.

- Steve
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://redfish.com/pipermail/friam_redfish.com/attachments/20060601/5e4acff2/attachment-0001.htm

[Martin C. Martin-2]

Hi,

The archives to this mailing list are available in one-month chunks, but
there doesn't seem to be any way to download the entire archive or
search the entire thing.

This "split into months" seems to be standard for mailman, and I suspect
it was chosen many years ago.  Now that both disk space and bandwidth
are so much bigger, does it make sense to switch to one big archive
file?  Is this possible/easy to do with mailman?

Thanks,
Martin

[Carl Tollander]

I second that, the archives are getting to be of a size where a
search/indexing facility would be
very useful.

Carl

Martin C. Martin wrote:

> Hi,
>
> The archives to this mailing list are available in one-month chunks, but
> there doesn't seem to be any way to download the entire archive or
> search the entire thing.
>
> This "split into months" seems to be standard for mailman, and I suspect
> it was chosen many years ago.  Now that both disk space and bandwidth
> are so much bigger, does it make sense to switch to one big archive
> file?  Is this possible/easy to do with mailman?
>
> Thanks,
> Martin
>
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org
>
>
>
>  

... [show rest of quote]

[Carlos Gershenson]

On 01 Jun 2006, at 21:25, Carl Tollander wrote:

> I second that, the archives are getting to be of a size where a
> search/indexing facility would be
> very useful.

Maybe to put a Google searchbox only for the archive directories?
I'm not sure if Google allows to do directory-only searches. If not,  
I'm sure http://www.master.com allows for it (also free).

Best regards,

     Carlos Gershenson...
     Centrum Leo Apostel, Vrije Universiteit Brussel
     Krijgskundestraat 33. B-1160 Brussels, Belgium
     http://homepages.vub.ac.be/~cgershen/

   ?Tendencies tend to change...?

[Stephen Guerin]

I briefly looked into this a few times in the past.

Here's a relevant FAQ from Mailman:
http://www.python.org/cgi-bin/faqw-mm.py?req=show&file=faq01.011.htp

Of the options listed there, I'll experiment with adding friam traffic to
mail-archive.com and let you know how it goes.

-Steve

 

> -----Original Message-----
> From: Martin C. Martin [mailto:martin at martincmartin.com]
> Sent: Thursday, June 01, 2006 8:30 PM
> To: The Friday Morning Applied Complexity Coffee Group
> Subject: [FRIAM] Searchable archives?
>
> Hi,
>
> The archives to this mailing list are available in one-month
> chunks, but there doesn't seem to be any way to download the
> entire archive or search the entire thing.
>
> This "split into months" seems to be standard for mailman,
> and I suspect it was chosen many years ago.  Now that both
> disk space and bandwidth are so much bigger, does it make
> sense to switch to one big archive file?  Is this
> possible/easy to do with mailman?
>
> Thanks,
> Martin
>
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org
>
>

... [show rest of quote]

[Owen Densmore]

How about GMane:
   http://gmane.org/
As far as my experience, its by far the best of the mail/web/search  
interfaces.

     -- Owen

Owen Densmore
http://backspaces.net - http://redfish.com - http://friam.org


On Jun 1, 2006, at 9:55 PM, Stephen Guerin wrote:

> I briefly looked into this a few times in the past.
>
> Here's a relevant FAQ from Mailman:
> http://www.python.org/cgi-bin/faqw-mm.py?req=show&file=faq01.011.htp
>
> Of the options listed there, I'll experiment with adding friam  
> traffic to
> mail-archive.com and let you know how it goes.
>
> -Steve
>
>
>
>> -----Original Message-----
>> From: Martin C. Martin [mailto:martin at martincmartin.com]
>> Sent: Thursday, June 01, 2006 8:30 PM
>> To: The Friday Morning Applied Complexity Coffee Group
>> Subject: [FRIAM] Searchable archives?
>>
>> Hi,
>>
>> The archives to this mailing list are available in one-month
>> chunks, but there doesn't seem to be any way to download the
>> entire archive or search the entire thing.
>>
>> This "split into months" seems to be standard for mailman,
>> and I suspect it was chosen many years ago.  Now that both
>> disk space and bandwidth are so much bigger, does it make
>> sense to switch to one big archive file?  Is this
>> possible/easy to do with mailman?
>>
>> Thanks,
>> Martin
>>
>>
>> ============================================================
>> FRIAM Applied Complexity Group listserv
>> Meets Fridays 9a-11:30 at cafe at St. John's College
>> lectures, archives, unsubscribe, maps at http://www.friam.org
>>
>>
>
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org

... [show rest of quote]

[Stephen Guerin]

Archives and RSS feeds of Friam are now available at gmane.org.
http://tinyurl.com/zwjl4

-Steve

> -----Original Message-----
> From: Owen Densmore [mailto:owen at backspaces.net]
> Sent: Thursday, June 01, 2006 10:18 PM
> To: The Friday Morning Applied Complexity Coffee Group
> Subject: Re: [FRIAM] Searchable archives?
>
> How about GMane:
>    http://gmane.org/
> As far as my experience, its by far the best of the
> mail/web/search interfaces.
>
>      -- Owen
>
> Owen Densmore
> http://backspaces.net - http://redfish.com - http://friam.org
>
>
> On Jun 1, 2006, at 9:55 PM, Stephen Guerin wrote:
>
> > I briefly looked into this a few times in the past.
> >
> > Here's a relevant FAQ from Mailman:
> > http://www.python.org/cgi-bin/faqw-mm.py?req=show&file=faq01.011.htp
> >
> > Of the options listed there, I'll experiment with adding
> friam traffic
> > to mail-archive.com and let you know how it goes.
> >
> > -Steve
> >
> >
> >
> >> -----Original Message-----
> >> From: Martin C. Martin [mailto:martin at martincmartin.com]
> >> Sent: Thursday, June 01, 2006 8:30 PM
> >> To: The Friday Morning Applied Complexity Coffee Group
> >> Subject: [FRIAM] Searchable archives?
> >>
> >> Hi,
> >>
> >> The archives to this mailing list are available in
> one-month chunks,
> >> but there doesn't seem to be any way to download the
> entire archive
> >> or search the entire thing.
> >>
> >> This "split into months" seems to be standard for mailman, and I
> >> suspect it was chosen many years ago.  Now that both disk
> space and
> >> bandwidth are so much bigger, does it make sense to switch
> to one big
> >> archive file?  Is this possible/easy to do with mailman?
> >>
> >> Thanks,
> >> Martin
> >>
> >>
> >> ============================================================
> >> FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at
> >> cafe at St. John's College lectures, archives,
> unsubscribe, maps at
> >> http://www.friam.org
> >>
> >>
> >
> >
> > ============================================================
> > FRIAM Applied Complexity Group listserv Meets Fridays
> 9a-11:30 at cafe
> > at St. John's College lectures, archives, unsubscribe, maps at
> > http://www.friam.org
>
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org
>
>

... [show rest of quote]