donn@utah-gr.UUCP (Donn Seeley) (10/22/84)
I did a little study of truncated paths and I think I have found two connections which seem to be the sources. Of 23 articles with truncated 'Path:' lines in four newsgroups, 15 passed through the connection genrad->wjh12 and the remainder passed through the connection hplabs->hao. The fact that makes these two connections suspicious is that in every one of these articles, there is a terminal path fragment beginning with either 'genrad!...' or 'hplabs!...' that is exactly 128 bytes long. No other power-of-two truncation applied to the same set of paths gets a result that consistently lies on a machine-to-machine boundary in the path. I have found paths going through genrad and hplabs which do not show this truncation; all the examples I have failed to go through wjh12 or hao, however. This suggests that the truncation occurs on genrad or hplabs when the article is batched for shipment to wjh12 or hao, or it happens in an unbatcher or in rnews on wjh12 and hao. It may be a coincidence, but both wjh12 and hao are recorded as using early September versions of 2.10.2 news... The guilty parties can probably be uncovered by inspection of news articles in the uucp queues on the affected machines. Could the news administrators on genrad, wjh12, hplabs and hao please take a look? Donn Seeley (decvax|hplabs|ihnp4|seismo)!utah-cs!donn