Earlier this week I mentioned the Datalex cache and how everyone is now pushing the relative merits of their own caching technology. I know that EB2 used to tout the strength of their cache (now owned by Sabre I assume they still do) and then yesterday I read some interesting stuff from ITA on caches.
In December 2008 ITA filed a patent application with the title Method and apparatus for providing availability of airline seats. The abstract appears below:
A computer program product, method and system for producing seat availability information for a mode of travel such as airline travel produce a prediction of availability of a seat in accordance with an availability query. The prediction is used in place of making an actual query to an airline or other travel mode availability system.
That patent had the names Carl Demarcken and Gregory Galperin on it, but then in September 2009 it appears they refiled the patent with the same title, but a new inventor added (David Baggett), and a rewritten abstract.
An availability system used for a travel planning system includes a cache having entries of availability information of seats for a mode of transportation. The system includes a cache manager that manages entry information in the cache so that information in the cache is correct, current, complete or otherwise as useful as possible. The cache manager determines when a stored answer is stale and, if a stored answer is stale, sends an availability query to a source of availability information
One of the perils of writing here is that occasionally I am really pushing the limits of my expertise – clearly patent law is something that is not my forte, but it is an interesting area nonetheless. Probably the most interesting piece of corporate promotional fluff that I am mailed regularly is the newsletter I receive from law firm Griffith Hack who specialize in intellectual property; unfortunately reading this once a quarter hardly makes me an expert!
ITA cache patent application US2009234682 (A1)
I’ve lifted a few interesting slabs of text from the patent application, but it really was a very long document and I only had time to skim read it.
[0043] Set forth below are several cache management strategies. In practice multiple strategies can be mixed together and executed simultaneously to meet multiple goals at once. The availability system uses data sources which asynchronously notify a travel planning system 10 of schedule changes or updates; the cache manager 150 can track these notifications and use the information contained therein to further guide cache insertion and deletion. For instance if the cache manager 150 receives a schedule change notification that a flight has been canceled, it can remove all entries relating to that flight from its cache. Similarly, if it receives notification that a flight has been added, it can create entries related to that flight and place them on lists to be added or modified in the cache. Finally, there are data sources such as so-called “AVS messages” which asynchronously notify the system of availability data of certain flights; the cache manager 150 can treat those just as it would responses directly from the availability data sources, and enter that data into the appropriate entries in the cache if appropriate, add entries to the cache, or simply ignore the messages
and then
[0057] Another version of this system replaces the process of gathering access counts in real time with a predictor of that value. One way of making such a predictor is to model one from historical data as follows: the above system is run to gather a database of lists of entries and access counts: instead of deleting the lists as prescribed above, the list is collected in a database for later processing. When the database is large enough, corresponding entries (e.g., “all US Air flights out of BOS before 11:00 am” or “US6309 11DEC BOS-LGA 10:00″) are averaged to get one mean predicted value for each entry in the list. A list of these averages is then used rather than constructed lists described above. While entries referring to specific absolute dates are unlikely to generalize and should largely be omitted from the compiled list, entries making reference to relative dates (such as “one week from now”) are likely to be very useful.
They really are trying to cover a lot of bases with this application, but why not give it a try. I mentioned in the earlier piece on Datalex the challenge for airline buyers to compare one caching technology over another, but an even more difficult challenge may be enforcing a patent on caching technology given that almost by definition a cache is a mysterious black box that only the creators can see inside of. Maybe the patent application is more of a marketing exercize to solve the first problem of differentiation in the eyes of buyers. I’m not exactly sure how one cache creator would know how another cache was working just by looking at the output, but just as I’m not an expert on patents, neither have I ever actually tried to build a cache, so this is just one man’s hypothesizing.
Whilst reading up on caches I came across another very interesting and curious patent application that is also airline related. Nothing to do with direct channels, but it definitely ticks the innovation box for creative and radical ideas. It was called System and Method of Control of the Terrestrial Climate and its Protection against Warming and Climatic Catastrophes Caused by Warming such as Hurricanes and I’ll leave you with the opening line of the abstract to ponder over the Easter break:
This system of the control and protection of the terrestrial climate relies mainly on civilian airlines burning (preferably price-subsidized) sun-shading (sun-blocking/sun-reflective) fuels in the high levels of the atmosphere in order to reduce the intensity of the solar radiation reaching the Earth’s surface.
April 1, 2010 at 9:52 am
Martin,
Congratulations for your posts on the cache issue, they are excellent.
On topic: there is no workaround for the real time interrogation of the inventory loaded in a GDS. You may build
very complex cache systems that are trying to simulate the avilability received from GDS, but you will end up with an error sooner or later as the accuracy of the system cannot technically reach 100%; more, due to the specific issues that occur, you will never be close to this figure. The predictive approach is equally risky, as long as it is based on predictios and
patterns recorded in the past – the reality is complex and unexpected inventory chanes can occur at any point in time, AVS messages can get queued and delayed, all of which which can contradict exactly those estimated patterns and will not match the real inventory of the air carrier you are trying to sell.
I was searching this morning for a RT flight OTP-MAD-LIS on iberia.com website and at payment time, when tried to enter the personal details of the passengers, the following error occured: Se ha encontrado una discrepancia entre el precio mostrado y la disponibilidad real de plazas en los vuelos seleccionados. Por favor cambie la selección e inténtelo de nuevo.(15079) (A discrepancy between the price shown and the real seat availabilty has occurred; please change your selection and try again (15079) ). This is very clear – the cache is severly out of sync since I could even reproduce the issue. Like I said before – there is no workaround for the GDS real time interrogation.
April 1, 2010 at 4:40 pm
Elvis, I know a lot of airline employees read this blog, but quite rare to get a comment from one; always an honour to get something from someone actually running an airline website.