Andrew Patterson wrote:
>> Experience from classic email - where orders of magnitude more messages are
>> processed - tells us that polling an account as a rule takes much less than a
>> second, and that it is indeed absolutely acceptable - even preferable.
> 
> Yes, but that's polling one account (or maybe 5 accounts for those
> with lots of accounts) using POP.. the actual delivery of the email
> to my mailbox is not done with polling but by the sending mail
> server pushing it to my account using SMTP.. I think Ross' point was that it
> is impractical to poll every medical facility in australia looking
> for pertinent results, much the same way I don't poll every
> POP server on the internet hunting for my email..

Sure, but my question was "Why is it necessary to poll every path and
imaging provider?" Surely there is only a need to poll the much more
limited number of providers to who the GP(s) in a practice have actually
referred patients/specimens. If a lab forwards a specimen to another
lab, then the results would need to be made available at the original
lab for collection from that lab's Web service.

> I agree though that we should learn from how email has done
> things - in particular, in SMTP the sender always takes
> responsibility for the delivery of the email - so it will keep
> attempting to send messages until it reaches timeout, all the
> while informing back to the originator if it is having problems. Why
> couldn't a path lab system keep attempting to initiate a connection
> to the GP system "web service" every 10 minutes until it successfully
> hands off the result?? Surely that would handle the unreliability of
> the australian internet without resorting to hub and spoke models?

The underlying assumption behind Ross's model, and also in the NeHTA
documents, is that GPs aren't up to running 24x7 servers which host Web
services. I have several thoughts on that:

1) Web services can be easily implement using packaged software on a
standardised Web server at the edge of each practice. That implies that
each practice has a fixed IP address (very shortsighted of the
broadband-for-health people to have ignored the suggestion that this be
a requirement - although yes, I know that dynamic-DNS solutions are
possible, but they are complex and often unreliable).

2) 24x7 operation is not required. The GP's Web service only needs to
run while the GPs are there at their practice (either physically or
virtually). As Andrew points out, automatic retries by the sender until
the message is able to be delivered solves the problem of a GP's Web
server being switched off for the weekend, doesn't it?

3) Alternatively, other organisations can run web services on behalf of
practices, and the practices can poll those Web services. Which is
really the model which NEHTA is pushing, I think. The critical thing is
that there is complete interoperability between Web services, so that
Divisions of GPs, ISPs and entrepreneurs and others can set them up if
they wish - remembering that all the content they are handling is
encrypted. The more of these, the better, as it removes major points of
failure. The Web services specs from NEHTA need to allow for each
practice to use more than one Web services provider to represent them on
the Web. Thus, there needs to be a primary and a secondary URL specified
for every request/referral for delivery of results/reports. The GP's
practice polls both the primary and secondary Web service providers, or
perhaps polls the secondary provider if the primary provider is down.
Clearly there are details to be worked out, and NEHTA needs to do that
ASAP if it wants its model to fly. Then it needs to sponsor a consortium
of individual Web service providers (not just the current messaging
providers) to demonstrate that this model works and to discover all the
flaws that no-one anticipated, and to test scalability, reliability and
resilience to point failures. All that testing might cost a few million,
but NEHTA now has the cash to be able to finance that sort of thing.
Commissioning a suitable group to write an open sourced reference
implementation of such a Web services server would cost perhaps a few
hundred thousand (including documentation and testing) and would
kick-start the process and encourage lots of redundancy by allowing
Divisns of GP, ISPS, even state health authorities to run Web services,
for themselves and on behalf of GPs. Some GP practices may run their own
as well. The more Web service providers there are, the better.

Then we would have a really sound, reliable, scalable, and open system
of health messaging/Web services in which market forces (due to the
openness of the infrastructure) quick force the prices down to the
lowest sustainable level (which in my view should be very low,
particularly when such an approach allows a "RAID" - a redundant array
of inexpensive delivery/collection [Web services].

Of course, all of this is still predicated on a pervasive PKI, isn't it?
Does NEHTA think that HeSA cuts the mustard in this respect, I wonder?
Given that a national provider index is planned in the next three years,
a national PKI intimately hooked up to the national provider index makes
enormous sense. NEHTA has, from memory, $30-40 million to implement
this. So let's have the national health PKI we ought to have had from
the outset. How about it, NEHTA?

Tim C

_______________________________________________
Gpcg_talk mailing list
[email protected]
http://ozdocit.org/cgi-bin/mailman/listinfo/gpcg_talk

Reply via email to