Date: Fri, 23 Aug 1996 23:03:44 +0200 (MET DST) From: Stefan Esser <se@zpr.uni-koeln.de> To: "Rodney W. Grimes" <rgrimes@GndRsh.aac.dev.com> Cc: se@zpr.uni-koeln.de (Stefan Esser), michaelv@HeadCandy.com, jas@flyingfox.com, freebsd-hardware@FreeBSD.ORG Subject: Re: SCSI again: Asus SC200 vs. Adaptec 2940 Message-ID: <199608232103.XAA22840@x14.mi.uni-koeln.de> In-Reply-To: <199608212007.NAA10653@GndRsh.aac.dev.com> References: <199608211913.VAA04328@x14.mi.uni-koeln.de> <199608212007.NAA10653@GndRsh.aac.dev.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Rodney W. Grimes writes: > ... > > > >I do have to say that I have an order of magnitude more trust in the > > > >aha2940 driver, and on my ability to get a bug fixed when I need it > > > >(Hi Justin! Thanks for the _GREAT_ support!!) by sending the broken > > > >hardware to Justin and giving him some time to work on it. > > > > Well, that really would help ... > > Why did I never receive a single piece of broken hardware to be able > > to reproduce a problem somebody sees with a single disk, tape or CDROM > > drive. Reading your reply it seems that you took my mail very personally ... That was not what I intended. I know your point of view, and while I doubt that I ever could help you in case of a technical problem (while I was able to debug failures with many other people quite effectively), my mail was in no way meant to bring up this point again. > Simple economics, it cost me less than $25.00 to overnight a drive and > controller combination down to Justin, $50.00 round trip as I usually > have him bill the return shipment to my accounts. It would cost me > on the order of $87.00 each way to ship to you, plus there is the whole > export/import/VAT and other taxes situation. > > It is very hard to get past VAT as you have to deal with temporary export/ > import paper work and all sorts of other oddities. It also takes me 5 > minutes to prep a shipment to Justin, it would take close to an hour to > prep an international temporary export. Well, the case that made me give up trying to offer support to you was the failure of a cheap CD-ROM drive (the Chinon 525), which most probably cost you less than $87 and thus would not have justified all this paper work for the temporary export :) > I make very little money selling SC-200 cards at $75.00 a piece there > just isn't much margin in them. Yes, sure. But you sell SCSI based systems at a reasonable price because such a cheap (and fast and well designed) SCSI card exists. > > This makes me wonder (again), whether I really should spend $1000 > > on the necessary hardware, and time worth some $10000 or more (if I > > spent it on paid work) just to allow other people to use the latest > > 53c8xx variants (Ultra-WIDE), who then complain about a NCR driver > > bug, whenever some VM system tuning introduces temporal instability :) > > Don't make me laugh... I've probably invested well into the 6 digits > of time and money into FreeBSD in one way or another. I am sure David, > John, Jordan, Poul, Julian, and a fist full of others are in the same > situation. No, I was not talking about the time and money I spent. I never counted hours, but I might try to get to some estimate. You might even beat me by an order of magnitude. That's not my point at all. I'm currently thinking about spending three weeks of vacation on all the cleanup required in the driver to make it fully support the more recent 53c8xx chips. And this will require me to buy a new system with an Ultra-Wide NCR controller and an appropriate drive. I'm not taking the price of the system into account, just what it costs to order those additional UW components, which I don't actually have any use for, except for driver testing. And its just no fun to invest these three weeks of time, which I might as well spend on some sunny island, if all I can hope for is that after the worst bugs introduced by the changes (there will be some, it will be a rewrite of major parts of the driver) have been found, people will go back to the normal level of NCR driver bashing. This driver has for quite some time (in late 1994) been the fastest in FreeBSD. It had for quite some time been the only driver that actually used Tagged Command Queues by default. Both of these made it much more likely for race conditions and locking problems in the kernel to show up, since the time required to start a command and the number of commands in operation at a time could be higher than with other controllers. There was quite some time correlations between large changes to the VM code and supposed NCR driver instability. And I invested tens to hundreds of hours into looking for errors in the driver that vanished as soon as other parts of the kernel had stabilized again. There have been real bugs (and real nasty bugs :) in the driver, and there most probably are still some more (in which driver aren't any ?). And I'm very interested in further improving the quality of the code. But I found that I can't do this besides my current job (there are only so many hours to a day ...) and I'm really wondering whether it is a good idea to spend all of my vacation on this. It's as simple as this ... Regards, STefan
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199608232103.XAA22840>