At PubSub, we’re beginning to put together a live feed of
earthquake and tsunami information in the hope that by making this information
more widely and rapidly disseminated, we’ll be able to prevent at least a
tiny amount of the troubles such as recently happened in the Indian Ocean. In
collecting this data, we’re not simply polling the USGS RSS feeds –
doing that would introduce too much latency in the process and such high
latency or delays would be unacceptable for this kind of data. What we’re
doing is connecting to live “push” feeds of data and then executing
on our servers the complex and messy “clean-up”, validation and
reformatting processes that are required. The instant we have good data, we push
it out to our subscribers using “Atom over XMPP/PubSub” and, as
always, insert the latest messages into our subscriber’s personal Atom
and/or RSS files. This service will be made available for at least experimental
use in a few days after we’ve completed some testing. We currently expect
to provide both traditional PubSub subscription service as well as a “telnet
feed” of raw data that other service providers will be able to use to
build their own systems. We think it is appropriate to feed this kind of data over PubSub and in
Atom files but before we get too far down this road, I’d like to get some
input from the Atom community about the format we’re thinking about using.
A sample Atom 0.3 entry from the current system is provided below. Please take
a look at it, provide whatever comments you consider appropriate and also
consider my questions below: 1. The
first thing to note is that the content is xml, not text or XHTML. The thinking
here is that we really want to be able to provide “data” to end
users since it is likely that the best way to use this data is to have some
kind of application (like a mapper, etc.) that would process the data. It is, I
think, less interesting to provide text. However, some folk will want a machine
generated textual version of this message… Since Atom only allows us one
content element, there is no way to do both data and text – or, am I
missing something? (Note: I do not want to convert this XML to XHTML…) 2. I’ve
been thinking that I might include an atom:summary element that contained a
textual version of the XML data carried in the content field. However, it
appears that most aggregators today display either the summary or the content
but don’t display both... 3. I’ve
provided a ‘rel=”alternate”’ link that points to a USGS
generated HTML page describing the event. Is it legitimate to point to an “alternate”
that I haven’t created myself? 4. The
title field provides some reasonably compact user data summarizing the key data
for the event. Would that be sufficient as a “textual representation”? 5. What
should a client do when presented with this data? Clearly, there are all sorts
of interesting things that one can do if you know the format and build a custom
app. However, it would be nice to allow “dumb” clients to do
something useful without having to force the format to
lowest-common-denominator (i.e. HTML or XHTML). I think what I’d like to
do is provide a pointer to an XSL-template in some standard manner. Then, a
client could fetch the style sheet and generate something for display. Is this
reasonable? If so, where would I identify the style sheet? 6. This
kind of data is really sensitive… I would very much like to be able to use
a Digital signature to ensure that we can repudiate any fake messages that
might get generated by spammers and other idiots as well as to repudiate mangled
versions of the data that might get produced down-stream of us. The subject of
signing Atom entries doesn’t seem to be getting much attention. Can we
wake this one up again? I really do not think that signing should be considered
something “outside the core” and subject to individual or
proprietary extensions. I look forward to your comments and assistance in getting
this system working – hopefully, long before the next disaster in which
it might be useful. bob wyman Note: The message below was published long after the original event.
The reason for the delay is that this message adds to information previously
published – much closer to the time of the event. (i.e. the “AddOn”
element with the links to the waveforms was not available when the event was
first published.) If a tsunami warning was issued, it would also appear as an “AddOn”. <entry> <modified>2005-01-08T16:49:18-05:00</modified> <issued>2005-01-08T16:49:18-05:00</issued> <id><id>tag:pubsub.com,2005:EQ:ci14117540</id></id> <title>M 1.8 SOUTHERN CALIFORNIA 2005-01-08T21:00:21.000Z</title> <link rel="alternate">http://earthquake.usgs.gov/recenteqs/Quakes/ci14117540.htm</link> <content type='text/xml' xmlns='http://pubsub.com/xmlns'> <Message> <topic>Earthquakes</topic> <publisher>PubSub Earthquake Publisher</publisher> <messageID>14117540</messageID> <revisionID>2</revisionID> <publication-date>2005-01-08T16:49:18-0500</publication-date> <contentType>EQMessage</contentType> <content> <Earthquake MsgNumber="1411754002"
MsgVersion="1.0" > <DataMessage Action=""
TimeReceived="2005-01-08T16:49:18-0500"> <Identifier EventIDKey="ci14117540"
DataSource="CI" Version="1"/> <Event Type="Earthquake" Version="1"
MsgTypeCode="E " Latitude="34.6391"
Longitude="-118.573" Depth="8.3" NumPhases="27"
MinDistance="5" AzimuthalGap="57.6" RMSTimeError="0.28"
HorizontalError="0.5" VerticalError="1.8"
Time="2005-01-08T21:00:21.000Z" LocationMethod="H"
Verified="N"> <Magnitude Value="1.8" Type="L"
NumStations="16" MagError="0.3"/> </Event> <AddOn Type="Waveform_CI" Version="1"
Link="http://pasadena.wr.usgs.gov/review_eventfiles/14117540.html"
Description="Waveforms"></AddOn> </DataMessage> </Earthquake> </content> </Message> </content> </entry> |
- Re: Please Review: Dissemination o... Bob Wyman
- Re: Please Review: Disseminat... Antone Roundy
- Re: Please Review: Disseminat... Robert Sayre
- RE: Please Review: Disseminat... Bob Wyman
- RE: Please Review: Disseminat... Bob Wyman
- RE: Please Review: Dissem... Ziv Caspi
- Re: Please Review: Di... Danny Ayers
- Re: Please Review: Di... Eric Brunner-Williams in Portland Maine
- Re: Please Review... James Snell
- RE: Please R... Bob Wyman
- Re: Plea... Peter Saint-Andre