Well I've been busily hacking away at my Sembrowser ... it's almost
functional.  Here are it's capabilities todate:

RDF/XML -> Sem (working has some bugs)
Semenglish -> Sem (working need to extend language to (subordinate clauses)
and anonomous nodes )
Sem ->  HTML Browser window (working)
Select Context (Contexts implemented similar to Grahm's paper.)

If your interested in the data structure of sem check out
[1] http://robustai.net/mentography/sempointers.gif

And how I implemented context
[2] http://robustai.net/mentography/contexttables.gif

This is working sufficient to get a view of where it is going....

The first thing that comes to mind is that contexts become very confining
and compartmentalizing.  If I say [Seth is Great] in context A, then
navigate to context B and say [Seth is Stupid] .. its like I'm in a
different world .. the two are not connected .. which is good and bad.  For
this to be interesting we're going to need to be able to do much more mixing
and matching and combining and refining of temporary contexts.  We need some
more theoretical work on how the context browser changes and combines and
filters contexts.

The next thing that comes to mind is that we need some more theoretical
thinking of how a personal sem interacts with a (sem web wiki).  It's easy
for the sem to just project its sem in HTML files on the web ... what's not
so easy is how the world will relate to that (sem web).  Here are some
ideas/questions:

* Each page of a sem web should contain an embedded RDF/XML file of its
contents ... perhaps there should be a icon on the page indicating that this
is the case.  That way a surfer can just acquire the metadata knowledge from
any page by reading the URL into there application.
* If someone on the web makes entries to a (sem web wiki) page, how does
that data flow back to a personal sem to which the page is associated ?
* How does one update the context associated with a (sem web wiki) page and
automatically update that page without destroying everybody else's comments
?
* How do we write Semenglish in a (sem web wiki) ?  I think we need some
QUICKWEB type tags designed to input Semenglish (and perhaps N3) to a wiki
page.

... well I thought i'd just leave you all with some ideas and questions
before I went back into hibernation.  I'll be setting the project on the
back burner until I finish some stuff for SpeakToMe ... then i'll get back
into it  in a couple of weeks.  Hopefully, then,  I'll be able to make a
download of an .exe file of the Sembrowser for people to play with.

.... until then ... find where this semantic web thingy is at ... ok?
Seth

From: "Sean B. Palmer" <sean@...>

>      http://purl.org/net/sbp/
>      - A Homepage of Sean B. Palmer

Congrats, and for the record see the semtran on ThisMessage.

But what is the use of the data uri in the subject?
data:text/plain;charset=UTF-8,http%3A%2F%2Fpurl%2Eorg%2Fnet%2Fsbp%2F

What do you all think of a bot that would cruise the lists we frequent, cull
the semtrans from the emails, and publish them to a (sem web wiki), that
would project a RSS feed of the changes to the greater semantic cloud?

semtran: language=Semenglish
Sean
     hasHomePage http://purl.org/net/sbp/ ;
     name "Sean B. Palmer";
     mbox mailto:sean@... .
Seth
     (wants to show you) (Quads Diagram);
     (likes the sound of) (sem web wiki).
(Quads Diagram)
     url http://robustai.net/mentography/quads.gif ;
     description "Shows the relationship between quads and sem spaces";
     comment " 'label' is to sem as 'uri' is to RDF".
(sem web wiki)
     description "A sem projected into a  navgatable web wiki".

Persistence of HTTP URIs depends on three crucial factors:-

      1) The astuteness of the creator.
      2) The permanence of the domain.
      3) The stability of the DNS system.

Regarding my own personal domains, I realised recently that although I
have full control over 1), I have no control whatsoever over 3), and
very little over 2). Therefore, in order that my homepage (the most
important resource that I maintain, for it links to everything else
that I do...) persists for just a bit longer, I have decided to set up
a PURL for it:-

      http://purl.org/net/sbp/
      - A Homepage of Sean B. Palmer

I hope that the concept of that URI shall persist, and that it will
save me the problem of when I can no longer afford to run
"infomesh.net". If you need to refer to the concept of my homepage,
please do so using the new PURL and not the old "infomesh.net" URI
(although I will still continue to maintain that for as long as
possible).

The reason I've sent it to SEM-DEV is so that this message itself may
be persistently archived, and maybe some of the follow-up replies
(bah!) can be directed into the domain of "what's a URI in the SEM
context?".

Note the amended email signature:-

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://purl.org/net/sbp/> .

Sean B. Palmer <sean@...> wrote:

> Aaron, can we have a graph merging interface on CWM at SWAG? Something
> like uri1=, uri2=, etc. so that we can merge SEMs. c.f.
> http://www.w3.org/2000/10/swap/Examples

I don't know how to write the HTML, but I've added this functionality. Just
add more &uri=... to the URL, like so:

http://purl.org/swag/cwm?uri=http%3A%2F%2Fwww.w3.org%2F2000%2F10%2Fswap%2Fte
st%2Fmeet%2Fblue.n3&uri=http://www.w3.org/2000/10/swap/test/meet/red.n3&uri=
http://www.w3.org/2000/10/swap/test/meet/green.n3

--
[ Aaron Swartz | me@... | http://www.aaronsw.com ]

Aaron, can we have a graph merging interface on CWM at SWAG? Something
like uri1=, uri2=, etc. so that we can merge SEMs. c.f.
http://www.w3.org/2000/10/swap/Examples

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://infomesh.net/sbp/> .

Hi JFD

> You do well to suggest that CG should wed to RDF; I'll look into that.

Sombody recently showed that you could transform CG to RDF.  They published
on the RDF-IG.  If my Sembrowser was together I could point you right at it.

> I don't use CGs directly, though.
> I'm interested in text analysis into
> 1 tagged text, with argumentation represented in XML

Well I can think of nothing better than to represent your augmentation of
text in RDF ... it was designed exactly for that purpose.  Though CG may
have been designed to explicitidly represent conceptual dependancies (like
tertuary relationships like give) it is not as mainstreme as the W3C's RDF.
If people, like yourself, work with us to form consensus of how to represent
natural text; then we will all get somewhere.

> 2 a convenient rep outside of the text

Yes i think i understand.  I started my current quest about '93 by posting
that same question on comp.ai.nat-lang ... i didnt get a good answer then ..
just blank stares.  Now I think we at least have a start.

> I have been busy teeching/translating/engineering, and bearing the
> Canadian winter, but I'll have to give a push to my project.

What is your project?

> What do you work on yourself, the high-level framework only, or also
> conversion from some type of input? (as suggested by your domain
> name!)

I work on the Sembrowser project .. right now I'm implementing Contexts in
SEM.  It's scoped in the two graphs below, we have also started a mailing
list SEM-DEV.

... references that I haven't already provided follow in Semenglish for you
and the sake of the SethBot.

language: Semenglish
RDF-IG
     http://lists.w3.org/Archives/Public/www-rdf-interest/ .
SEM-DEV
     http://groups.yahoo.com/group/sem-dev .
Sembrowser
     description "allows a user to browse the semantic web";
     seeAlso
         http://robustai.net/mentography/cmap_vs_sem.gif ,
         http://robustai.net/mentography/SemBrowser.gif ,
         http://groups.yahoo.com/group/sem-dev .
SEM
     acronymOf "SEmantic Memory" .
JFD
     mbox mailto:mrojas@...
     name "JF Delannoy";
     homePage http://www.site.uottawa.ca/~delannoy ;
     affiliation "University of Ottawa, RES International";
     roles "Adjunct Professor, Senior knowledge engineer".

From: Sean B. Palmer mentioned enormous problems:

1) Multiple formats

Sem helps us here.  If there were 4 formats without sem, we would need 16
different transformations to cover the space.  With sem we need only 8
because each format need only have a input to sem and output from  sem.

2) Formats embedded in other formats (screen scraping)

A Better term might be (data mining).  There is a wealth of data just
sitting in our inboxes and bookmarks.  I'm surprised more people have not
taken a hack at it.

3) Lack of RDF data

I doubt that this will be a problem in the long run.  As soon as we get our
(data mining) tools together we should have a glut of RDF data.  The
inevitable problem will be more how do you filter that glut to the context
of your immediate concern.

Seth

----- Original Message -----

From: Aaron Swartz

To: sem-dev@yahoogroups.com

Sent: Thursday, March 15, 2001 5:33 AM

Subject: [sem-dev] Re: Coughing Up SEM

Seth Russell <seth@...> wrote:

> think, you in particular Aaron, have worked, designed, and related to, the
> external environment almost exclusively; and so tend to ignore the extra
> coherence that is achievable in a closed world.

I don't think so -- read my post more carefully. Didn't I just talk about
the coherence of a local memory?

--
[ Aaron Swartz | me@... | http://www.aaronsw.com ]



To unsubscribe from this group, send an email to:
sem-dev-unsubscribe@yahoogroups.com



Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

> Well see, that's the difference between a (semantic cloud)
> where what you say about not always being able to vance
> is definitely True, and a (semantic memory) in which, by
> definition and design, we can always vance.

Thus the problem is how do we get information from a
not-always-vancable, messy semantic cloud into our contextualized and
coherent semantic memories? At the moment we have to contend with
three things:-

    1) Multiple formats
    2) Formats embedded in other formats (screen scraping)
    3) Lack of RDF data

That's three enourmous problems that need to be faced.

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://infomesh.net/sbp/> .

Seth Russell <seth@...> wrote:

> think, you in particular Aaron, have worked, designed, and related to, the
> external environment almost exclusively; and so tend to ignore the extra
> coherence that is achievable in a closed world.

I don't think so -- read my post more carefully. Didn't I just talk about
the coherence of a local memory?

--
[ Aaron Swartz | me@... | http://www.aaronsw.com ]

--- Seth Russell <seth@...> wrote:
> Sean B. Palmer <sean@...> wrote:
> >From: Aaron Swartz
> >Maybe just post them to the rdfnet.allbots group?
>
> That a good place for semtrans, but too few people
> know about it and less
> use it.   I think we need a higher visibility entry
> point to the (semantic
> cloud) ... one that takes very little programming to
> read from and no
> programming to write to.   Any ideas ?
>

It occurred to me last night while reading over the N3
Primer that the organizations hosting on-line
dictionaries (:dictionary a :book) would be interested
in creating and hosting RDF vocabularies. In fact, I
think it is inevitable.

I've always felt a tinge of guilt when visiting an
on-line dictionary like http://www.m-w.com to look up
a definition, as if I was stealing something. Now I
realize that these organizations are making the
mistake of confusing analog objects with digital ones.
A better long-term model would be for them to get
involved with the creation of quality RDF
vocabularies.

John Hebert

__________________________________________________
Do You Yahoo!?
Yahoo! Auctions - Buy the things you want at great prices.
http://auctions.yahoo.com/

From: Aaron Swartz

To: sem-dev@yahoogroups.com

Sent: Wednesday, March 14, 2001 10:02 PM

Subject: [sem-dev] Re: Coughing Up SEM

Seth Russell <seth@...> wrote:

> ... hmmm... i guess it would work ... your uri is vancable by you ... mine
> by me ... mine by you only if i grant you access.

I just depends how you vance. This is the argument against always vancing --
a smart SemBrowser would look the URI up locally before it vanced.

--
[ Aaron Swartz | me@... | http://www.aaronsw.com ]



To unsubscribe from this group, send an email to:
sem-dev-unsubscribe@yahoogroups.com



Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

Seth Russell <seth@...> wrote:

> ... hmmm... i guess it would work ... your uri is vancable by you ... mine
> by me ... mine by you only if i grant you access.

I just depends how you vance. This is the argument against always vancing --
a smart SemBrowser would look the URI up locally before it vanced.

--
[ Aaron Swartz | me@... | http://www.aaronsw.com ]

From: Aaron Swartz

Seth Russell <seth@...> wrote:

> We need a schema to identify particular statements within sems.

>A schema is very different than a URI scheme, which I'm not convinced we
>need.

Nor am I, yet.

> >So if there is a statment in my sem that your sem knows about
>> then you could refer to it with a URI. [...] I think we do need a new
> >uri schema for that.

>Why? A URI scheme (oversimplified) implies a way to access data. With your
>URI scheme, what system would I use to access the data you describe. It
>seems like HTTP to me. What benefits are gained from this new scheme?

Well if the sem were connected to a http server, then http would work as
well ... i suppose the uri of a statement might look like

   http://robustai.net/bots/SethBot/?stid=2001.3.14.1234

.. so if you have access to my bot then you could type that into your
browser and you would see my statement in it's context.

But what of your bots memory of that statement which might look something
like:

[http://robustai.net/bots/SethBot/?stid=2001.3.14.1234
  Sean english:is english:Great]
     a english:exaggeration.

... that memory about exaggeration might have a uri in AaronBot like

http://aaronsw.com/bots/AaronBot/?stid=2002.4.1.22

... hmmm... i guess it would work ... your uri is vancable by you ... mine
by me ... mine by you only if i grant you access.

... just some more bozometry befoe i totally crash ...
Seth

Seth Russell <seth@...> wrote:

> We need a schema to identify particular statements within sems.

A schema is very different than a URI scheme, which I'm not convinced we
need.

> So if there is a statment in my sem that your sem knows about
> then you could refer to it with a URI. [...] I think we do need a new
> uri schema for that.

Why? A URI scheme (oversimplified) implies a way to access data. With your
URI scheme, what system would I use to access the data you describe. It
seems like HTTP to me. What benefits are gained from this new scheme?

--
[ Aaron Swartz | me@... | http://www.aaronsw.com ]

From: Aaron Swartz

>I'm saying we need a crawler like this to help us create/discover the
>Semantic Cloud.

I agree :)

Seth

From: Sean B. Palmer

>Therefore, I conclude that the "sem:" scheme is a wild-goose chase,
>and hereby decide to concentrate my efforts elsewhere. Now, where did
>I put my drink?

I can think of a use for "sem:", though it's quite different than what you
were talking about.  We need a schema to identify particular statements
within sems.  So if there is a statment in my sem that your sem knows about
then you could refer to it with a URI.  For example, the SethBot might have
a statement in its memory [sem://robustai.net/bots/SethBot#2001.03.14.12345
Sean english:is english:great.] that your bot knows about ... then you could
say [sem://robustai.net/bots/SethBot#2001.03.14.12345  english:is log:True].
Note the first statement is a quadrapole and the second a triple which makes
a statement about a statement in SethBot's memory.  I think we do need a new
uri schema for that.   And if  you pasted the uri
sem://robustai.net/bots/SethBot#2001.03.14.12345  into the address space of
your browser, it should show you that statement (in your bot) and it's
surrounding context.  If I typed the sem: into SethBot it would show me the
statement as it appears in SethBot's memory.

Another example that assums that I can obtain reified statement uri (as
above) from the Swag Dictionary....

Seth
     thinks [sem://webns.net/mvcb/Prefix#2001.02.02.123 (is better named)
NsPrefix.].

Seth Russell <seth@...> wrote:

>>> What's a scutter?  That URL doesn't help much.
>> Dan's web crawler for RDF.
> I gathered that, but what does it gather, what format and what syntax does
> it expect, what does it do with it, how do you included your web page in the
> scutter?    Are you saying that this is a high visibility entry point to the
> semantic cloud ?

It gathers all sorts of http://www.w3.org/1999/02/22-rdf-syntax-ns# and
stores it in a database which is used to generate http://www.rdfweb.org/ and
you get your webpage submitted by having some else reference it using
rdfs:seeAlso or by sending it to genid:danbri.

I'm saying we need a crawler like this to help us create/discover the
Semantic Cloud.

--
[ Aaron Swartz | me@... | http://www.aaronsw.com ]

From: Aaron Swartz

Seth Russell <seth@...> wrote:

>> Can we start a scutter for SEM?
>> http://rdfweb.org/~pldab/rdfweb/scutterplan.txt
> What's a scutter?  That URL doesn't help much.

>Dan's web crawler for RDF.

I gathered that, but what does it gather, what format and what syntax does
it expect, what does it do with it, how do you included your web page in the
scutter?    Are you saying that this is a high visibility entry point to the
semantic cloud ?

Seth

Seth Russell <seth@...> wrote:

>> Can we start a scutter for SEM?
>> http://rdfweb.org/~pldab/rdfweb/scutterplan.txt
> What's a scutter?  That URL doesn't help much.

Dan's web crawler for RDF.

--
[ Aaron Swartz | me@... | http://www.aaronsw.com ]

Sean B. Palmer <sean@...> wrote:

> Thus, by "cough up your SEM", I mean "cough up your real world data".
> There must be tons of little scraps of data that we all write down
> from day to day... is it really going to be that difficult to write it
> in N3 and post it to the Web?

Yes let's do it!  Eventually we'll find this so called Semantic thingy or
bust ....

From: Aaron Swartz

>Can we start a scutter for SEM?
>http://rdfweb.org/~pldab/rdfweb/scutterplan.txt

What's a scutter?  That URL doesn't help much.

>Maybe just post them to the rdfnet.allbots group?

That a good place for semtrans, but too few people know about it and less
use it.   I think we need a higher visibility entry point to the (semantic
cloud) ... one that takes very little programming to read from and no
programming to write to.   Any ideas ?

One idea I had was to data mine the email's that flow on the groups that SM
people frequent.  Just look for semtrans.  Then perhaps we could find a bot
that would collect them and publish them as perhaps a RSS channel or
something.  Thing is that RSS would need to be able to include any kind of
RDF ... not sure it can yet.  Maybe the bot could convert to Semenglish so
that it could be included in RDF.

I think people writing N3 or Semenglish at the bottom ef their emails to
list their citings, or define their terms, or publisize their sites, or
describe themselves is something that is a natural desire.  Most of the
software tools are in place already ... all we need is the bot to do the
data mining .. and a place to publish in a popular distribution media.  I
guess blogging is the other way ... anybody got a good (sem wicki) we could
all agree to use?

Some of these terms are defined in the Semtran below .... these definitions
obviously need some work .. plese don't be put off by my continual gyrations
of terms and syntax in Semenglish ... that's the way i always work .. i dink
around with things till they finally look right .. lots of penemento.

language: Semenglish
ThisMessage
     context SEM;
     informs {
SEM
     acronymOf "SEmantic Memory";
     composedOf (RDF statements);
     seeAlso http://robustai.net/mentography/SemBrowser.gif .
Semtran
     dfinition "communications in any RDF type languages to the (Semantic
Cloud)".
(Semantic Cloud)
     definition "all RDF statements communicate in any fashion on the
Internet".
(Semantic Web)
     definition "all RDF statements that are accessable by URL on the
Internet";
     partOf (Semantic Cloud).
Semenglish
     definition "Semantic English - English with unambiguous semantics";
     url http://robustai.net/mentography/semenglish.html ;
     objective "Semenglish is all about bringing to the foreground the
natural language that we all know, love, and understand; and at the same
time not sacrificing the precision that machine agents require one bit.  It
is designed to be not only the input from humans to (semantic memory), but
also the output from (semantic memory) to humans.".
(sem wiki)
     definition "a wiki that is backed by a sem data base and where it is
easy to enter triples as such".
Seth
     (still looking for programming help with) Sembrowser;
     (was working today on implementing) Context;
     (says go buy something from) http://speaktomecatalog.com .
}.

> Can we start a scutter for SEM?

That'd be a good idea, but would need co-ordination more with Seth
than myself.

As for the "sem:" scheme...

Say you want to browse N3. You want to be able to click a link and
have it processed by CWM. Now, this means that your browser has to
recognize the resource as Notation3... which is currently impossible.
I suggested sem: to denote anything of an RDF nature (O.K., it still
doesn't solve the problem but...) which meant I could use sem: links
to do whatever processing.

However, compare this to images. They don't need to have their own
scheme: they are delivered using whatever URL and the browser sniffs
the content. What if CWM were integrated into IE5.5 and then N3 files
all had the same content type (text/n3+rdf)? I wouldn't be asking for
a new URI scheme then would I...

Therefore, I conclude that the "sem:" scheme is a wild-goose chase,
and hereby decide to concentrate my efforts elsewhere. Now, where did
I put my drink?

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://infomesh.net/sbp/> .

Sean B. Palmer <sean@...> wrote:

> Thus, by "cough up your SEM", I mean "cough up your real world data".
> There must be tons of little scraps of data that we all write down
> from day to day... is it really going to be that difficult to write it
> in N3 and post it to the Web?

Can we start a scutter for SEM?

http://rdfweb.org/~pldab/rdfweb/scutterplan.txt

Or maybe just post them to the rdfnet.allbots group?

--
[ Aaron Swartz | me@... | http://www.aaronsw.com ]

So, how is SEM development these days then? Seth's experimentation
with "SWAG as seen by SethBot" have been receiving some attention, and
I have been working on "sem:" and CWM a little more... The first thing
on the agenda SEM wise is that everyone should be coughing up their
SEM.

Merging Notation3 is quite easy; just a case of feeding it through...
and if you have rules in there, you can even get CWM to -think about
it... but what is there to merge/think about? Seth showed that you get
easily compile huge lists of triples and display them all neatly in
XHTML; but at the moment the only terms in there are stuff like
ontologies, no real world data.

Thus, by "cough up your SEM", I mean "cough up your real world data".
There must be tons of little scraps of data that we all write down
from day to day... is it really going to be that difficult to write it
in N3 and post it to the Web?

That's where my proprietery URI protocol comes in handy. If we start
exchanging sem: URIs between ourselves, we can process them. What I
was thinking of is this:-

1) A sem: URI is put on the list, something like
"sem:http%2A%3A%3Ainfomesh.net%3Asem%3A".
2) Using Protozilla [1], it is possible to make that URI actually *do*
something.
3) In this case, I'd like to a) Bookmark the URL, b) retrieve, store,
and mege the content at the end of the URL.

I guess I should mention XHTML =XSLT=> N3 and so on here, but I won't.
Data tranmutability is becoming a bit of a pain, quite frankly... at
least the two main formats are easily converted 'twixt each other, but
personally I can't wait for more semEnglish tools and Seth's
databases.

[1] Thanks to Aaron, we have a reference to the great
http://protozilla.mozdev.org/ Protozilla is a Netscape 6 deeley that
lets you add your own protocols! While I'm not too sure about the
freedom that gives, at least I've been able to hack it to manage sem:
redirects in NS6. Thanks again to Aaron.

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://infomesh.net/sbp/> .

> Could you send me the exact things you are typeing into the
registry?

At the moment I am up to the stage where I'm asking it to simply open
things in Notepad, and it won't even do that... but maybe I can go
onto the nxt step.. right, I'll paste some code in between the XML
"<reg>" tags that you should be able to type into your registry. I'm
using regedit.exe (which should be in the WINDOWS folder) to put this
in at the moment, because then you can see exactly what you're doing.
BTW, the "EditValue" is a binary thing...

<reg>
[HKEY_CLASSES_ROOT]
[sem]
(Default) = "URL:SEM Protocol"
EditFlags = "02 00 00 00"
"URL Protocol" = ""
[DefaultIcon]
(Default) = "C:\WINDOWS\SYSTEM\url.dll,0"
[shell]
[open]
[command]
(Default) = ""C:\Program Files\Python20\python.exe" cwm.py %1 -think"
</reg>

Substitute your python path for the one above. This entry doesn't work
at all for me in IE5.5, wheras at least the "note:" one that I did at
least says "cannot find file". *sigh*

> Apparently this should work, if not, then we should know
> the reason why.

Yes: it should work, but it doesn't, and I have no idea why not.

> i don't know how to access a COM server without writing some kind
> of program .. do you?

I'm afraid not. I'm terrible at object oriented programming - I've
never really had the time or the tools to learn. That's becoming quite
a big handicap... I have the ideas but I can't put them into practise.

> sem:purpose=ask&language=Semenglish&context=SEM&
> url=http://robustai.net/public/SethBot
>
> ??

That'd be O.K., we could use anything we wanted really; "whatever fits
the tools best". We're in the very early stages of doing the tools, so
whatever works is nothing short of a triumph!

> It's interesting, seems like our knowledge in this area is
> somewhat complementary

Yes, it's very useful! We both come up with ideas that seem to further
the work that the other does... very isoDynamic :-)

> So,  are you saying that a URI that is a <sem:something>
> is doing what

Well, it's more like "sem:metadata,data-address" in that the first bit
might be able to tell our SEM browsers what to do with the URL. For
example, you might just want to view a semTran, or you might want to
load it into your own SEM, or process it against your SEM, split it
from your SEM, get your SEM to recognize it, and so on. It could also
contain format instructions, like data: has "type", but this'd have
triples type, like N3, semEnglish, tab-delimited, and so on. The MIME
type would be up to the data-adress used, so an HTTP URL could have
any MIME type relayed over it.

> I was thinking of something more like SOAP

I don't know all that much about SOAP... something about HTTP transfer
and XML... could you enlighten me?

> > Let's hope so! I thought about it last night after I experimented
with
> > "freenet" [2].  [...]
> Your way ahead of me :)

Not really... if you were, then you'd be behind the rest of the world,
because I certainly am, and I see you as being two steps ahead of the
SW community. No-one else even discusses contexts of SEMs, and yet
they are surely going to be a driving part of the whole
infrastructure... so that confuses me a bit. Maybe other people just
haven't had that "aha!" moment yet?

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://infomesh.net/sbp/> .

From: "Sean B. Palmer" <sean@...>

> > >      sem:http://infomesh.net/myN3/
> > [...]
> > Lot's of luck ... I wouldn't know where to start ...
>
> Well, it's got something to do with editing the registy to just add a
> protocol... but I can't quite get it to work. It keeps saying that
> there's "nothing on the disk" when I know very well that there is. The
> example I gave above was a highly simplified version of what I'm
> looking to achieve - I want it to be a bit like the "data:" protocol
> that has some information in it about the nature and purpose of the
> URL (that'll be the next step!).

Could you send me the exact things you are typeing into the registry?  Then
I could take a hack at it.  Apparently this should work, if not, then we
should know the reason why.

> Hey, that's a good step! I have no idea how to use Jason's COM thing
> yet, so I should have a look into it really...

I think you would need to be in some programming environment like VB, of VFB
or C++ ... i don't know how to access a COM server without writing some kind
of program .. do you?

>
> > But I have a question about your uri structure above ... it's got
> two
> > colons ... so you are actually nesting uri schema .... can you do
> that?
>
> I think so: for example, I'm pretty sure that some URNs have colons in
> them. I'd like the structure of "sem:" to be something like:-
>
>      sem:[type]=[value];[type]=[value],URL
>
> Where [type] can be information such as "purpose", "name", "format",
> and so on, and where a URL is anything that a processor can vance upon
> or process to get some SEM.

Ok, I think I see where you are going.

Incidentally Semenglish would have a problem with the embedded ";" and ","

.... couldn't we find some way to eliminate those .... maybe use & like ....

sem:purpose=ask&language=Semenglish&context=SEM&url=http://robustai.net/publ
ic/SethBot

??

> > Why not just use http as the schema?
>
> Do you mean use the HTTP format, or the HTTP scheme? i.e. use HTTP to
> get out SEMs, or the HTTP URI string method to refer to a computer
> name and then use whatever format you can find? If you mean the
> former, well I expect/hope that people will use lots of different ways
> to vance upon their SEMs, "gopher:", "http:", "telnet:", "https:",
> "ftp:", "data:", "freenet:", and so on, so I wanted to show some way
> that what we're getting at the end of them is a "SEM". There were
> three options:-
>
>    1. Use content negotiation (i.e. MIME types)
>    2. Use some kind of suffix
>    3. Use some kind of new scheme
>
> 1 was out of the question because we all use widely varying formats
> for our triples, and you can't tell that just because something is
> text/xml that it contains triples. Likewise, is text/plain in
> Notation3, or just random garbage? 2 was also out of the question
> because it would be confused as part of the URI string itself.. so I
> decided to go with a new URI scheme.

I tend to agree.  It's interesting, seems like our knowledge in this area is
somewhat complementary ...you are much better versed in these particlar
areas than am I.

So,  are you saying that a URI that is a <sem:something>  is doing what?  Is
it supposed to name someting?  Or is it a kind of a semtran?   If it is a
semtran, then I think it needs more structure than a uri .. or rather, were
we to code all we need to do in a stemtran in a uri, then we will be having
lots of troubels.  I was thinking of something more like SOAP or Dave's
super simple precursor to SOAP and all of those just use the simple http
get/put protocal that we already have.  I think that is what i was
suggesting.  Does that make sense?

> If you mean that we can use the HTTP structure, i.e. create an
> entirely new kind of protocol, then that would be a good idea, but
> we'd have to hack up some kind of "SEM server" that could send data in
> a SEM format... no easy task.

No, literally,  I meat just use http get and put to implement communicating
semtrans between sems.

> Let's hope so! I thought about it last night after I experimented with
> "freenet" [2]. It's a way of storing infomation on the Internet a bit
> like the Web, but it's completely anonymous, and I wondered how that
> could be covered with SEMs... so I set this up with the intention of
> sending my crazy ideas here, and then releasing it to you all when I
> had something substantive.

Your way ahead of me :)

Seth

> >      sem:http://infomesh.net/myN3/
> [...]
> Lot's of luck ... I wouldn't know where to start ...

Well, it's got something to do with editing the registy to just add a
protocol... but I can't quite get it to work. It keeps saying that
there's "nothing on the disk" when I know very well that there is. The
example I gave above was a highly simplified version of what I'm
looking to achieve - I want it to be a bit like the "data:" protocol
that has some information in it about the nature and purpose of the
URL (that'll be the next step!).

There's some information about adding new schemes on a W3C addressing
page [1], but it just wont work. It crashed my browser a couple of
times...

> When I enter a URI in the read address box of my SethBot (name
> of my sem) it will read/process that URI.

Hey, that's a good step! I have no idea how to use Jason's COM thing
yet, so I should have a look into it really...

> But I have a question about your uri structure above ... it's got
two
> colons ... so you are actually nesting uri schema .... can you do
that?

I think so: for example, I'm pretty sure that some URNs have colons in
them. I'd like the structure of "sem:" to be something like:-

      sem:[type]=[value];[type]=[value],URL

Where [type] can be information such as "purpose", "name", "format",
and so on, and where a URL is anything that a processor can vance upon
or process to get some SEM.

> Why not just use http as the schema?

Do you mean use the HTTP format, or the HTTP scheme? i.e. use HTTP to
get out SEMs, or the HTTP URI string method to refer to a computer
name and then use whatever format you can find? If you mean the
former, well I expect/hope that people will use lots of different ways
to vance upon their SEMs, "gopher:", "http:", "telnet:", "https:",
"ftp:", "data:", "freenet:", and so on, so I wanted to show some way
that what we're getting at the end of them is a "SEM". There were
three options:-

    1. Use content negotiation (i.e. MIME types)
    2. Use some kind of suffix
    3. Use some kind of new scheme

1 was out of the question because we all use widely varying formats
for our triples, and you can't tell that just because something is
text/xml that it contains triples. Likewise, is text/plain in
Notation3, or just random garbage? 2 was also out of the question
because it would be confused as part of the URI string itself.. so I
decided to go with a new URI scheme.

The problem with creating a new scheme is that people are going to
moan about it :-) Aaron commented just the other day that making new
URI schemes on a whim is a bad idea... but I don't consider this to be
a "whim".

If you mean that we can use the HTTP structure, i.e. create an
entirely new kind of protocol, then that would be a good idea, but
we'd have to hack up some kind of "SEM server" that could send data in
a SEM format... no easy task.

> Thanks for creating this group, it should be a big help with getting
> sems going :))

Let's hope so! I thought about it last night after I experimented with
"freenet" [2]. It's a way of storing infomation on the Internet a bit
like the Web, but it's completely anonymous, and I wondered how that
could be covered with SEMs... so I set this up with the intention of
sending my crazy ideas here, and then releasing it to you all when I
had something substantive.

I quickly realised that I couldn't do this without your help :-)

[1] http://www.w3.org/Addressing/schemes.html#hack-schemes
[2] http://freenetproject.org/

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://infomesh.net/sbp/> .

From: "Sean B. Palmer" <sean@...>


> Hi Seth,
> I sent a couple of messages and then realised that this group would be
> impossible without you on it. The current thing I'm trying to do is to
> get it so that when I type:-
>
>      sem:http://infomesh.net/myN3/
>
> into IE5.5, it processes the N3 on CWM. Phew! The couple of messages
> I've sent already are at:-
> http://groups.yahoo.com/group/sem-dev/messages

Lot's of luck ... I wouldn't know where to start ... but i guess it would be
possible .. even desirable.

When I enter a URI in the read address box of my SethBot (name of my sem) it
will read/process that URI.  I have this working now with RDF/XML via
Jason's parser.

But I have a question about your uri structure above ... it's got two colons
... so you are actually nesting uri schema .... can you do that? ... is
there any precedance for this that you can provide a url of ?

In response to another of your posts, I agree, we need a URI format for our
sems.   Why not just use http as the schema?   For example, I'm on the @Home
cabel, so I could setup a http server on my work computer here.  I forget
the URL that they gave me but lets say it was 206.235.992.123 (which it
isn't) ... so I could connect SethBot to

     http://206.235.992.123/publicAccess/SethBot/

This then would be the URI , the URL, and the machine name for SethBot ...
in other words:

language: Semenglish

SethBot
     uri   http://206.235.992.123/publicAccess/SethBot/ ;
     url   http://206.235.992.123/publicAccess/SethBot/ .

language: English

What do you think?

Thanks for creating this group, it should be a big help with getting sems
going :))

Seth

> Because there will be potentially hundreds of thousands of SEMs,
> and much interchange and discussion, we need to establish protocols
> and agreements to ensure that its development remains interoperable
> and pushes it "to its full potential" :-)

As such, we will be wanting to do the following:-

    1. Have public SEMs
    2. Have private SEMs
    3. Have group SEMs
    4. Have SEMs with precise editing authorizations
    5. Have SEMs referable by URIs
    6. Have SEMs vancable by URIs
    7. Refer to individual triples within SEMs
    8. Merge SEMs
    9. Split SEMs
    10. Evolve SEMs
    11. Swap, share, and develop SEMs

As such, I think that all SEMs will have to have a URI if they are to
be useful. What kind of URIs could SEMs have? Well they'd be URLs:-

      http://infomesh.net/mySEM/
      ftp://infomesh.net/mySEM/
      data:,URI1%20URI2%20URI3.

The last one poses an interesting problem: how to differentiate
between a SEM and the name of a SEM. The last URI is a SEM in itself,
not a name for a SEM. Maybe (and this is dodgy reasoning) it would be
a good idea to have a "SEM space":-

sem:name=MySem;type=N3;purpose=address,http://infomesh.net/mySEM/

This would indicate that a resource is to be "got at" with the
protocol defined after the ",", but that it has this attached
metainformation (the stuff before the ","). The "sem:" scheme means a)
this is a SEM, b) this should be processed by my favourite SEM
processor.

I wonder if I can get it to work under Win98?

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://infomesh.net/sbp/> .

Welcome to sem-dev: SEM = SEmantic Memory.

This is a group that I have set up to (in the long-term) explore some
of the technical aspects of SEM, and (in the short term) to
investigate some architectural ideas that I wish to develop.

What is SEM? SEmantic Memory is a set of contextual triples in the
form URI1(URI2,URI3) that is used to provide some coherance when
downloading information from the Semantic Web.

Because there will be potentially hundreds of thousands of SEMs, and
much interchange and discussion, we need to establish protocols and
agreements to ensure that its development remains interoperable and
pushes it "to its full potential" :-)

Some of my short term ideas are going to be slightly controversial,
and hence input for the first phase of this group will be very
limited. In short, this is just a "test group" to see if such a
developmental process will be useful to the Semantic Web community.

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://infomesh.net/sbp/> .