Hard Launch Day 2

Some fun, some fear, but it all came out right in the end.

===========

I had put the XML output into its own printer definition, so that the users would do all their pagination and output , and then do the XML output at the end. We had planned to incorporate this into our final output printer definitions once every pagination had some CCI ads in it. We made the change this afternoon, copying the output into the XMLPaginationReport folder on the CCI server as well as a backup location. The initial output has a pagination status of 3 for all ads; this means "in progress." At 1am we run a cron job that takes the backup file and recopies it with the pagination status changed to 10 or "published" into the XMLPaginationReport folder again.

I had forgotten to set the "write ad list" parameter to X and put the file name in the ad list name area, so I was getting zero-length XML files. Gary fortunately found the problem and we got it all straightened out.

===========

21 CCI ads did not make it into the extraction. Because of the difficulty in creating them one by one if there is no matching class and sort, I edited the classsrt.out file manually to add them in as display ads. Then we brought the eps files in one by one by pointing the ad records to the CCI folder containing the ads by ad name.

============

We had some confusion about the merge files. I have been using symbolic links to existing folders because of the difficulty of getting the CCI and Sii files together in the same folder. I pointed to the wrong folder for the Sii files, but it looked correct because an erroneous extraction from Sii had been made the previous week, so the files that had been there were one week old. I should have noticed the problem in the dates with the new merge list that comes up in ClassPag, but I didn't, and this cost us time before we noticed the problem. We were looking at duplicate ads and wondering why things didn't look right. Moral: always pay attention to the merge list and check the dates assiduously.

=============

We plan tomorrow to pull out the proofs from one week ago to make sure we aren't missing anything in our pagination "furniture". We copied our existing paginations but there have been some changes since then that could trip us up. Our paginators have good memories but it is better not to depend on that when checking to make sure everything is exactly right in the new pagination.

==============

I only had a couple of minutes where I actually lost my ability to think.

==============

-JoAnne

Hard Launch at Baltimore Sun

Our first day with the merge went reasonably well. Some notes:

==========

Our legacy system, Sii, had been modified to accept another field, the new classification field, e.g. the five-digit classification such as 40100, 40150, 40200 and so on for Employment, for example. The field is just a type-in field and is not validated against a table of known good classes. As a result we got a few bogus classes which caused problems in pagination. It was clear that we had bad classes because they appear in the class listing at the beginning of the pagination by number, rather than being inside the defined class super groupings with names (e.g. Auction, auto, etc.).

We addressed this in two ways: one, we went back and fixed the field in the affected ads, re-extracted, and paginated. Second, on an extraction which was running very late, I went in and modified the classsrt.out file manually to correct the erroneous class. However, there is no easy way to know if a keying error resulted in a classification that exists but is not correct for the ad until it is noticed or someone complains.

===========

The Death index process that we developed is sort of involved, and did not include all ads for some reason. SCS may be enlisted to help develop something better. It may involve parsing the ceps.out file to try to extract the first line with the deceased's name, or may involve having CCI output something we can use.

===========

Even one merged pagination took a lot of time. If we were to do this over again we might introduce only one at a time rather than three (Auto/Market, Deaths, Boats).

===========

Fonts have to be checked and rechecked at every CCI upgrade and change in procedure. You should do a full test all the way to output on any change. We discovered late that there are apparently two symbol fonts used on CCI, one in a dropdown, one in the font set. We had to install the second one (called just Symbol) quickly all along the proof and film output paths on the day of launch.

===========

Rather than trying to use the reports from SCS which never seem to have what we want, I have been creating a report generated from data extracted from the full-blown outext.txt file, imported into SQL server, then interpreted via Crystal Reports. This will soon be fully automated and we will be collecting daily data on all our paginations for future uses we might only imagine today. Given the current state of data availability from CCI, this might be an approach worthwhile for other properties.

===========

Test early and often! Courage!

============

-JoAnne