Re: interesting bug issue

[Seth Teller <seth@graphics.lcs.mit.edu>]

another idea:  as a by-product of the BMG processing pipeline,
we could compile a structured database of all named spaces on
campus, and make it available in a light-weight fashion as a
series of configurable menu picks, the completion of which
would produce a well-formed space name of our own design.

MIT's "whereis" service has a page something like this at:

   http://whereis.mit.edu/doc/locate.html

in our setting, one could be directed to a web page (or to
in-lined code in an existing event-creation page) presenting
a menu pick such as:

   building -> NE43

   [another button appears, with floors 0-10]

   floor -> 4

   [another button appears, with all 4th floor space names]

   room -> 423A

at which point the string NE43.4@4@423A@CONF is produced.

this turns the problem around, in that it generates only those
space names that we know how the handle.  the challenge of
course is to make the generation process so easy and widely
available that most people will use it.

Seth Teller wrote:
> 
> interesting.
> 
> why is "FLOOR" duplicated in our naming scheme?  (.FLOOR@FLOOR)
> 
> regarding name lookups:  one could imagine writing a parser for
> all space descriptions at MIT, but the space of names is finite,
> so instead i recommend creating a table that can be built up
> lazily (i.e., whenever a name is entered that does not already
> have an match in the table).  the table could simply be a
> collection of strings:
> 
>   "Grier A", "34-401A",
>   "Teal Room", "26-152",
>   etc.
> 
> so your code would be a (case- and whitespace-insensitive)
> string search, nothing fancy.
> 
> patrick, doesn't the LCS events server log an email address for
> the creator of each record?  you could have your code email a
> question to the creator whenever an unrecognized space designation
> was used.  of course, what would the question be exactly.  it
> could be "please specify the formal building and room number
> for XXX", where "XXX" is the human-specified location (e.g,
> "Grier Room").
> 
> but what if the specified location has no room number (e.g.,
> "Lobby 7", "the Dot") or has one but it is unknown to the event
> specifier (e.g., "classroom nearest to landmark abc")?
> 
> another issue is context:  "4th floor lounge" means one thing
> w.r.t. LCS and another w.r.t. other buildings.
> 
> the only thing i can think of is to direct the event specifier
> to a web page displaying the floorplan containing the event
> location (if we can determine it), and ask him or her to indicate
> the named space interactively, say with a mouse-click.  we could
> help out by highlighting the space outlines upon mouse-over.  is
> this relatively easy to do using our existing java API ?
> 
> other ideas ?
> 
> seth.
> 
> Patrick Nichols wrote:
> 
>> Gents --
>>
>>
>> * * *
>> Just an FYI about some interesting issues I'm hitting as I write
>> code that you might have some thoughts on.  I'm writing some code that
>> probes the LCS events database on the web (right now it is UNDESCRIBABLY
>> ghetto and inefficient, but it works... sort of) and I'm hitting an 
>> issue:
>> for my signs to work, rooms are uniquely identified using a global
>> namespace. The naming scheming Jason used (and hence I use) looks like:
>>
>>   NE43-218 -> NE43.2@2@18@OFF
>>
>> or generallly:
>>   [BUILDING NAME].[FLOOR]@[FLOOR]@[ROOM NUMBER]@[OFF|CONF]
>>
>> The issue is that in the LCS events database, most rooms are referred to
>> in their general usage at MIT (eg NE43-218) or as text descriptions ("the
>> 4th floor lounge").  I'm writing a lookup function for now that 
>> translates
>> rooms to their known namespace equivalent, but I'd appreciate it if 
>> any of
>> you code ninjas out there can think of a better way of doing things.
>>
>> -Patrick
>>
>>
> 
> 
>