This note defines the D44 archiving subproject. First, we justify this project from a Beam Physics Analysis perspective. We then layout the requirements, establish and briefly comment on the existing prototype. The D44 data is currently sutided in Main Control Room (MCR) and in off-line analysis. While we typically look over fairly recent period (a day or less) in MCR, we are often interested in establishing trends over much longer periods. Note also that the previous luminosity record was obtained in October 02 and the most recent record in March 03. It take long time to improve our performance. Typical studies are various lifetime vs beam intensities, loss profiles using the experiments counters, emittance growth and so forth. Finally, we do a very superficial job at store post mortem dumps. This is part due to the lack of expert analysts, in part due to the fact we do not have enough recorded data: significant lapses between similar anomalous store failures do occur. Requirements: 1. The ACNET data coming from all Tevatron instruments (FBI, Toroids, SBD, SyncLite, BLM, Tune Fitter, B0 and D0 loss counters, as well as other future instruments) will be recorded at a frequency of about 1 Hz, or at a lower frequency dictated by the output rate of the relevant front-end. In addition, critical machine parameters will also be recorded. 2. This data will be kept for at least 3 years. 3. The OSDA-D44 Java package will be extended to provide access to this data. Status of the prototype: A document has been written in September 02. (Document # The software is in CVS (osdaarchive). This D44archiver is currently running on stkendca8a, and could serve data if the SuperTable needs a rebuild. However, it fails to collect data occasionally. The access to the D44 loggers has been recently upgraded, we still have problems. The data on the stkendca8a is not compressed (the "gunzip" has not been implemented). The data is saved on tape although the procedure is not automated. We therefore need to review the prototype and decide how to implement the real thing before proceeding with a better implementation. The amount of man-power is probably in the range of three man-months.