SQL query to get UP and Down jobs ignoring subseqent dups

Here is the trouble I need to track system outages based on batched
jobs.

I have the following data in tables (headers)

Job, Start Date, Start Time,End Date, End Time


This is a batch schedule with jobs that make a system unavailable.

Now there are two types of batch jobs, a start and an end so we get
the following:
--job log table--
ENVSTAT,8/26/2007,08:17:12,8/26/2007,8:18:28
ENVSTPM,8/26/2007,19:25:46,8/26/2007,19:26:06
ENVSTAM,8/26/2007,19:51:13,8/26/2007,19:51:33
ENVSTPA,8/27/2007,09:23:09,8/27/2007,9:35:24
ENVSTPA,8/28/2007,16:00:41,8/28/2007,16:12:49
ENVSTPM,8/28/2007,19:15:18,8/28/2007,19:15:38
ENVSTAM,8/28/2007,20:19:04,8/28/2007,20:19:26
ENVSTAC,8/28/2007,20:23:46,8/28/2007,20:24:07
ENVSTPA,8/29/2007,00:06:20,8/29/2007,0:18:22
ENVSTA1,8/29/2007,01:05:33,8/29/2007,1:17:57
ENVSTPM,8/29/2007,19:12:34,8/29/2007,19:13:05
ENVSTAM,8/29/2007,19:55:31,8/29/2007,19:55:51
ENVSTPB,8/29/2007,20:12:00,8/29/2007,20:27:57
ENVSTAB,8/29/2007,20:42:18,8/29/2007,20:54:45
ENVSTPM,8/30/2007,19:16:49,8/30/2007,19:17:11
ENVSTAM,8/30/2007,19:50:03,8/30/2007,19:50:23
ENVSTPB,8/30/2007,20:36:10,8/30/2007,20:52:05
ENVSTAB,8/30/2007,21:11:34,8/30/2007,21:24:01
ENVSTPA,8/31/2007,03:59:12,8/31/2007,4:15:59
ENVSTA1,8/31/2007,05:42:03,8/31/2007,5:54:53
ENVSTPM,8/31/2007,19:10:33,8/31/2007,19:10:54
ENVSTAM,8/31/2007,19:41:28,8/31/2007,19:41:48
ENVSTPB,8/31/2007,19:41:59,8/31/2007,19:54:38
ENVSTAB,8/31/2007,20:13:28,8/31/2007,20:25:53
ENVSTPA,9/1/2007,00:46:02,9/1/2007,1:02:49
ENVSTA1,9/1/2007,02:23:23,9/1/2007,2:35:54
ENVSTPM,9/1/2007,19:37:23,9/1/2007,19:37:45
ENVSTAM,9/1/2007,21:14:19,9/1/2007,21:14:39

Now certain jobs will make the system unavailable (down jobs) and some
will bring them up. We need to ignore batches that do not bring
environments down (in the above example C and T jobs do nothing to
us.)

What I am after is the down time. I am going to store in a separate
table (jobdef) the up and down jobs for reference. The trick is they
are linked in pairs (STPM,STAM ; STPA, STA1 ; etc.) but can run
concurrently so if an M batch brings the environment down at say noon
to 1. But an A batch brings it down from 1:30 to 2 then the real down
time is from noon to 2 (See how this is tricky?) But if M brings
things down from noon to 1 and A brings is down from 2 to 4 that is a
new second down time.

// EXAMPLE
On the 26th the M Batch took the system down from 19:25:46 to 19:51:33

BUT
on the 31st the M batch took the system down at 19:10:33 but it
finally came back up at 20:25:53 (The B batch continued the outage).
The current way I'd record the outages is from 19:00 to 20:00
(rounding to 30 min blocks.) which is not accurate IMHO.



The question is this:

How do I get the start time and date of the FIRST down job, ignoring
subsequent down jobs (it's already down) and get the end time of the
FIRST up job ignoring subseqent up jobs (because it's already up.) but
taking into account overlapping down periods.

I currently do this by hand and it takes hours upon hours and I do it
in 15 min increments (not accurate but faster). I want "to the minute"
accuracy and no human contamination in recording the availability so I
am going to put this into a MYSQL database initially and make a report
via a web page rather then hacking this manually via Excel (we're
talking days worth of my life doing this.)

I believe in working smarter, no harder but I cannot for the life of
me figure out the SQL needed to pull this out for graphing.

The main reason is I compare the sum of the down time against a
schedule and report the difference but I have to show it graphically.

Here is my best psuedocode

FOR A GIVEN TIME PERIOD STARTING AT X AND ENDING AT Y

MAKE a TEMP RECORD with the START TIME of the FIRST DOWN JOB AND the
END TIME of the LAST UP JOB that went down DURING THE FIRST DOWN JOB.
// in short we have to take into account overlapping batches

In the end I should have just a list of downtime windows like
temp1,01/01/07,8:00AM.9:45PM

Oh and one last gotchya, given this is reasonably in realtime we have
to assume that the system is down to NOW() if we have a down job but
no start job at the end.

I am lost on this and have been trying to tinker with it for over a
year and haven't gotten very far. I need my weekends back, help! :)