20040929: ldm 6.1 - change in "-m" behavior

Hi Harry,

> To: address@hidden
> From: Harry Edmon <address@hidden>
> Subject: ldm 6.1 - change in behavior
> Organization: University of Washington
> Keywords: 200409282351.i8SNpxHP002118 LDM 6.1

The above message contained the following:

> I start my rpc.ldmd processes with a "-m 10800".  In the past if I 
> started the ldm with a new queue file it would get data 3 hours old.  I 
> just tried this with the new ldm, it seems to have done it for some of 
> the feeds, but not all.  Is this a bug?  The first 100 lines of my 
> ldmd.log are attached.

You've been relying on undocumented (or, at least, badly documented)
behavior.  The "-m max_latency" option specifies the maximum latency
that a downstream LDM 5 will accept before responding with a RECLASS
reply.  It also specifies how far back to request data-products
unless overridden by the "-o toffset" option (which must be less than
"max_latency").  The value used for this "backlog" request will be
adjusted, however, by the creation-time of the most recent data-product
in the product-queue that matches the selection criteria -- but only
if the creation-time is more recent than the current time minus
max_latency.

If the "-o toffset" option is used, then the backlog time isn't adjusted
by anything in the product-queue.

I know, this is horribly convoluted.  I tried to keep the original logic.

If you absolutely, positively must get data-products from 3 hours ago --
even if they're already in the product-queue -- then use the following:

    rpc.ldmd -m 10801 -o 10800 ...

Regards,
Steve Emmerson