I have yet to be convinced that there is a single, solitary case where
cancellation in transit would prove useful.
Here's one. I accidentally mail a 5 megabyte video when what I meant to
mail was an external reference to a 5 megabyte video. Cancellation in
transit would be quite useful here; save lots of bandwidth, disk, and cpu
on intermediate sites.
As it happens this is a case I've thought a lot about, albeit not in the
context of this particular service element. And I'm far from not convinced it
is really that useful. (This assumes that all the implementation problems are
solved, of course.) It may even be counterproductive.
People learn to adapt. Sure, somebody on a slow link might send a huge
message once, but hopefully they'll learn from that mistake and won't do it
again. In fact the ability to use up what's available and no more is
really quite remarkable.
Software engineers learn too, and if their users don't seem to be able to
control their usage user agent software will adapt and take a more active role
in preventing these mistakes.
Cancellation services, especially haphazardly implemented cancellation
services (which any new service is guaranteed to be) may end up interfering
with this process and therefore may be counterproductive.
Mind you, I'm still on the fence about this one. I just don't think it is
as clearly useful as you apparently do.
One thing that makes in-transit cancellation at least *interesting* is that
it mirrors one example in the paper world. If I FedEx someone a package, I
can have that package returned to me undelivered, so long as I call FedEx
early enough in the process. The Internet currently works much more like
the US Post Office, where you should be greatful if packages get delivered
at all, and they'll laugh in your face if you want one back.
There are examples of this in email as well.
Perhaps we have no choice but to be the USPS. I'd like to at least think
about whether we could be FedEx, though.
Could is one thing. Should is another. As I have already said, acceptance
of this model has a profound impact on our ownership model.
beneficial. I never said there weren't. However, this doesn't mean that
these
operations need to be part of the transport infrastructure.
I didn't say they did. "In transit" was just one of four states I listed,
and I never said that cancellation "needed" to work there; just that it
would be OK if it did.
I'm still not convinced that there's really any advantage to it working in
transit. However, I'm already quite convinced that there's a big downside to
insisting on the ability to do in transit cancellations, a downside we have
not even begun to seriously discuss.
Ned