Yesterday we took the reference generator offline for nearly an hour to carry out a code and database update. The goal was to provide better and more accurate citations, that goal has been achieved even though the site was offline more nearly an hour. We feel it’s worth and hopefully you will agree.

Our service is built by aggregating book information from several sources (you will see these sources being acknowledged when you do a search). Naturally each data provider uses their own format.  We have been dealing that with successfully for quite a while. However occasionally we find that the same provider has subtle differences for different items of data. For example data provider A might have one format format for Catch 22 but  slightly different format for Battle Cry.  We had planned for it and our software is capable of detecting many of these subtleties and dealing with them successfully. However as the site grows and the usage increases, we are discovering new gremlins that we need to address. And that’s exactly what we did yesterday.

Foor example, you might have seen certain results where the authors data of birth had been prefixed to his name or you have have seen extra punctuation at certain positions. You might also have seen extra characters next to the year of publication. All these were manifestation of the subtle difference in formats encountered for certain data items – the gremlins. We are not complaining because the data providers re doing a great service without charging a penny for it. It’s up to us to accommodate them. And that’s what we did through the code updates yesterday.

We need to carry out a minor update to the Harvard Reference generator so we will need to take the website off line today (Friday 26, 2010)  at 4:00 PM GTM/UTC. Yes it’s a a bit of a short notice but the downtime will be less than 10 minutes.

Last week we shut down our service for one hour to migrate the site to a more powerful server. We are already seeing the benefits of this  much needed improvement.  Book searches are now a lot faster than they were before. We have given it a further boost by improving the code that powers the site. To add icing to the cake we have improved the caching mechanism that makes some searches a lot faster than before. Speed is not the only improvement, results are now more relevant and accurate.

Deadlinedue generates Harvard References for books and web pages by aggregating data from several different sources. Data from all the sources are analyzed to and determine which result is the best match the criteria you specified (book title , ISBN or web link). Then we look at the ‘good ones’ again to find out if they contain the full set of information required to generate a Harvard Reference. If both conditions are fulfilled we generate the result and display it for you. Since users often enter just part of the title instead of a full title, there is more work for us!

One of our datasources happen to be the Library Of Congress . However this source was being used only about 25% of the time for title searches and only about 10% of the time for ISBN searches. We have improved our code so that the LOC data can now be utilized for each and every search. However that does not mean all the results you see originate from this source. Sometimes the other data providers may have more relevant or more detailed results and when that happens we use the best possible data set.