On Wed, Sep 21, 2022 at 8:59 AM Clem Cole <clemc@ccc.com> wrote:
below in blue ...

On Wed, Sep 21, 2022 at 9:47 AM Warner Losh <imp@bsdimp.com> wrote:

So, maybe it did happen, but I find no extant evidence of a copyright being
removed and replaced by Berkeley. If anything, once files started being marked
with a copyright notice, they seem to be retained over several releases and
on the 2BSD series where the code was merged into.

On Wed, Sep 21, 2022 at 6:36 AM Rob Pike <robpike@gmail.com> wrote:
It was a long time ago but it rankled at the time (and even came up in my Bell Labs interview): that copyright notice replaced whatever was there before. Code written and copyrighted by other institutions was absorbed into the Berkeley distribution and reattributed without credit. Joy told me later that the lawyers made them do it. He was probably telling the truth, but that didn't make it OK.

I think there are two different concepts that are getting mixed up here.  The legal term 'copyright' and historical term of 'provenance.'    I agree with Warner that I know of few if any cases where copyright was not maintained when it was in the code itself.  And as he points out, please grep through the archives and I think that will be found to hold true. 

But I also think Rob rankle comment is fair.  Joy and was noted for recognizing cool ideas and adding them into 'Berkeley UNIX.  The line at the time was he took ideas and 'peed on them to make them smell like Berkeley.' For example, 'Berkeley Joy Control' came from Kulp via Europe and MIT, the network stack famously started at BBN, and a lot of the support for limits and user controllers from Australia.   

Yes, the CSRG team did do a great deal of innovation as well as integration, but the line between the two was not always easy to see from the outside.  And I think developers outside of UCB sometimes felt (to use Rob's words) 'rankled' for CSRG getting credit for some of innovation that really belonged to others, because the CSRG team was the distribution vehicle.

That makes a lot of sense. When there was a name, it was preserved, but a huge amount of the sources had nothing at all in the source files to identify it. One big area of contribution was into the kernel where the options sometimes contained the name of where the code came from. In the 2BSD kernels we see eg TEXAS_AUTOBAUD, MENLO_OVLY, MENLO_KOV, MENLO_JCL, MPX_FILS, CGL_RTP and a bunch of UCB_ names. It's clear the non UCB ones came from elsewhere, but there's no info on where they came from (they are documented in setup.5 at least so I know what they are). So given the sparseness of the early marking for provenance, the coments make more sense and give a better timeframe to it.