Japan relies on Delphion to update its US Granted imag table (they get the rest of their text data from the US/PTO's free, weekly bib-only FTP feed). Delphion also gives Patolis US Application data (all tables, main, abst, claims, as well as imag). The US Granted images come in on Wednesday and it takes Sandy 'till Friday at least to process them. We FTP the US Applications bibliographic data from the US PTO on Thursday, but the images don't come in until Friday and it takes all weekend to process those (unless, as often is the case lately, it seems), there are media errors necessitating Delphion to request a replacement tape. On top of that, the USPTO oftentimes distributes occasional images in even later weeks. For example, the 2,726 images we do have for the week of 1-3-2002 (out of 2,728 -- two we never got), were loaded in four different weeks; 1,375 on 1-24-2002, 9 on 2-25-2002, 1 on 2-28-2002, 1,341 on 3-17-2002. Another example is US21056583A1, issued 12/27/2001 but the image didn't come 'till 2002-03-13. Also, when bundling up the data to be shipped to Patolis, collect the data on a week boundary so that Patolis can (re)load and (re)index on those boundaries. Do not bundle multiple weeks of data together. To update all this US data manually, follow these 3 easy steps. It should take about 10 minutes total. ------------------------------------------------------------------------------- 1) Collect the newly-added information from a Southbury database (or replica). From jasper@jasper, ssh -l inst1 dephds059 set -o vi cd Japio.US.Update cat last_run to see when things last ran successfully. This file contains two lines, e.g. LAST_DATE='10/31/2002' LAST_IMAG_TS='2002-11-04-13.16.32' LAST_DATE is the Thursday issue date of the last set of US Applications data we extracted. The script will extract the data for any weeks it finds after that LAST_DATE. LAST_TS is the timestamp of the last time we ran, used to extract new US image data from the imag table. Run the script to extract the US Application data from the following 6 tables, abst, assg, clas_icl, clas_xcl, invt, main, and prir. The load files will be tar'd and compressed into YYMMDD.tar.Z. getUSdata.sh If it ran successfully, the lines in last_dates will be updated for the next run. ------------------------------------------------------------------------------- 2) Now follow the directions the getUSdata.sh script gives you to ship that (those) tar.Z file(s) over to Patolis, and uncompress. For example, Insure the destination directory exists in Patolis. I've pre-made these directories up through December, 2003, but if it's after that, as jasper on ips06i, mkdir /dfs/convdata/pa030306 where 030424 is the Thursday day, in this example, April 24, 2003. Back on dephds059, scp1 -p 021107.tar.Z inst1@ips06i.ips4db2.com:/dfs/convdata/pa030306 The password is inst1_password Then back as jasper on ips06i, cd /dfs/convdata/pa030306 zcat *.tar.Z | tar -xvf - rm -f *.tar.Z ------------------------------------------------------------------------------- 3) Send a note to the following people, telling him that the weekly US Apps data has been sent and resides in /ips/convdata/paYYMMDD. "Shizuo Shibata" "Takayuki Nakada" "Takashi Ogawa" "Yuji Takashima" The easiest way is to use the usapps nickname from Lotus Notes. One of them will load the files into DB/2 using their normal importpat procedure, then index the data, again with their normal bibupdidx procedure. OOO RRR O O R R O O RRR O O R R OOO R R if you want to load and index the stuff yourself, login as ipsadmin on ips03i, week=021031 cd /dfs/convdata/pa$week /ips/bin/importpat pa$week > importpat.log Then as ipsadmin on ips05i, week=021031 cd /dfs/convdata/pa$week /ips/bin/bibupdidx pa$week > bibupdidx.log If you're going to do the loading and indexing for them, it's worthwhile to check your work AND the status of their US Apps Verity index. rcvdk /ips/coll/coll_usapps | grep finished s banana quit You want to insure that the last number in this line, Search update: finished (100%). Retrieved: 6(6)/102758. 102,758 in this example on 4-24-2002, matches the number of US Applications in the database. db2 connect to patent user ipsrun using inst1_password db2 "select count(*) from main where patn like 'US2%'" which also matches the count in Southbury db2 "select count(*) from main where patn like 'US2%' and datasrc='UPA'"