Talk:Coherent (operating system)

[untitled]

The article says Coherent could run on "most" x86 CPUs, and starts the list at the 286. I know, for a certain fact, that the original 1984 Coherent ran on plain old waity-waity-eights. -- Geo Swan 13:39, 28 July 2005 (UTC)

[untitled]

Coherent 4.x for the 386 and above did not have virtual memory or demand paging. Nor did it have TCP/IP in the kernel. — Preceding unsigned comment added by 70.20.242.107 (talk) 23:10, 14 January 2006‎ (UTC)

[untitled]

It's weird to first state something, and then have a 'Correction' in the same article (regarding the open-sourcing of Coherent) — Preceding unsigned comment added by 192.100.124.218 (talk) 06:33, 12 September 2006‎ (UTC)

More than four people involved?

I still have a Coherent manual (1994). On the first page of the preface it states "Coherent is the work of a large number of exceptionally talented people." Below this it lists some of the people involved. There are about 75 names listed. This version was a 386 port. Maybe four people created the first release.

I bought Coherent for about a $100 (list price) to use to help train staff in Unix. You could boot DOS or boot Coherent on the same PC. Coherent was very small. At the time we were using SCO Unix for Oracle servers. SCO was several thousand dollars a server. The Coherent manual (about 1200 pages) was a better Unix reference than the documentation that came with SCO and that is why I still have it.

In 1990, my reading was that there were only two commercial grade Unix's that ran on PCs. SCO and one made by Kodak (yep the camera people). I forget the name of Kodak's Unix. They sold it in the early 90's. SCO at this time was located in Santa Clara and not run by those people in Utah.

One foot in the grave...

Being a firm believer that more information is better than less I offer my perspective. I was a heavy Coherent user between 1993 and 1995 and did some development on the platform.

Coherent already had one foot in the grave between 1994 and 1995 when Linux began to be popular. There were two problems, a bug that started to bleed off customers slowly and a missing feature which dropped Coherent from consideration for anyone doing new development.

The first problem was the introduction of a disasterous bug in the ide disk driver. At that time many Coherent users were looking for an inexpensive means of having a UUCP news feed. If I recall correctly C-News was part of one of the CohWare packs. The genesis of the bug was the addition of multi-sector reads to the "at" driver which was done as a performance boost. The new code did not do robust error checking. The problem manifested itself when one block of a multisector transfer failed. Rather than invalidating the whole transfer the code improperly scribbled on the next few active blocks in the buffer cache. This resulted in data corruption of the buffer cache and eventually the disk. Eventually the kernel would read a plain data block when it was looking for either a directory or a part of the on-disk inode table. At that time the kernel would panic and depending on your disk i/o load fsck had very little chance of repairing your filesystem. People using Coherent for their news feeds on MFM/RLL/IDE disks were hardest hit by these problems and complained very loudly in the newsgroup. It took Mark Williams a long time to fix this problem and while it was being fixed many users migrated to Linux.

The second problem and the single largest missing feature was TCP/IP networking. Hindsight being 20-20, the decision to port X windows before adding networking to Coherent was the genesis. The dedication of resources to both build a BSD socket API and then port a very green XFree86 stole valuable development and support time from the project of getting TCP/IP networking added to the kernel. Between 1994 and 1995 the Mark Williams staff included no more than 15 developers. Had the decision been made to do TCP/IP networking first, the effort needed to build a version of the BSD sockets API would have been paid for as part of networking project and the XFree86 port would have been considerably easier both because it was on a more "standard" platform and because the XFree86 code would have been considerably more mature. Adding networking would have allowed Coherent to take part in the TCP/IP revolution that began to sweep through the computing industry in 1995. Without TCP/IP Coherent was left in the dust as Linux, *BSD and even Microsoft hopped on the runaway train that we now call the internet.

I don't know if timely attention to the "at" driver bug and networking could have saved Coherent. However, it is my guess that those two factors would have kept Coherent viable in the market until the people could buy Red Hat Linux. At the time Coherent was the cheapest choice in the closed source PC Unix market place. Also at that time Linux and BSD were radical choices and espousing their merits got you shunned inside of corporations where the big wheels needed the comfort factor of having someone to call if the system broke at 0400.

24.44.29.247 18:29, 18 December 2006 (UTC)

Image of Coherent handbook cover

I could provide a scan or a picture of the Coherent handbook cover, if there is some interest for it - got it laying around here (no idea though wich version). Don't know if that's a bit overkill for this article...

Sir diddymus (talk) 09:37, 24 December 2007 (UTC)

Fred Butzen didn't do it alone

The article says "The manual was entirely written by Fred Butzen and was praised because of its quality". In fact, I wrote much of the manual in the mid-1980s. Jhobson1 (talk) 16:22, 10 August 2010 (UTC)

BSD

Is there any evidence, that Coherent is BSD? If not, please remove category "Berkeley Software Distribution". If yes, please add references to sources, which prove that. --Uniwersalista (talk) 15:43, 29 July 2013 (UTC)

Fixed! Letdorf (talk) 19:41, 29 July 2013 (UTC).