This is a summary of the feedback received at usenix, both from the 'Birds Of a Feather' session, and the NetBSD exhibition booth. If anyone is interested in helping with any of these issues, has further comments, or would be willing to write a similar report for any future conferences, please let us know.

Chuck Cranor chuck@research.att.com gave a talk on the design and implementation of the new UVM Virtual Memory System. UVM is specifically designed to provide the I/O and IPC systems with a range of flexible data movement mechanisms, and improves virtual memory performance over BSD VM in traditional areas such as forking and pageout. UVM is the standard VM system in NetBSD as of version 1.4, and is being ported to OpenBSD. Concern about people leaving core featured quite highly in the BOF, and several people approached the exhibition booth with the same issue. Jason explained that people tended to leave core to concentrate on writing code, and that he himself had managed to get much more code written since leaving core. There still seemed to be a general concern. People want it. One problem is that early versions of some files contained 'tainted' code which under the USL agreement we are not permitted to make available. Rather than take the easy option of throwing away all history before a certain date we chose to painstakingly check every file and specifically remove any problem versions. Perry spent some significant time on this and we now have a clean tree. The next stage is to make this available via anoncvs. This is underway and we hope to make some announcements over the new few weeks. People would like to be able to make tax deductible donations back to the project, in particular some of those who have benefitted from using NetBSD based solutions. Some could be quite significant. It would also be good to have an 'online donations' page. Some people would be more than willing to hire developers for specific driver support, but we do not currently provide any easy way for them to contact such developers. A web page listing 'consultants for hire', and those looking for such would be a good start in this direction. This could include categories such as developers, sysadmins, and systems integrators. Work has started on such a page. A number of companies are already using NetBSD in embedded applications, and others are interested, primarily due to the emphasis on clean code design and wide platform portability. We should setup a 'tech-embed' mailing list, and try to put up information on the website. There is also interest in adding real-time support to the NetBSD kernel. One particular aspect that appeals to some companies is the BSD licence. This allows them to use NetBSD without being obliged to make their source changes available. We obviously prefer them to do so, but agree it should be their choice. One option is to contribute back fixes and enhancement to the general code, while keeping specific sections confidential. We have a large number of active projects, and even a set of goals for the 1.5 release, but we do not make this information easily available on the website. People are more interested in writing code than telling people about it. There is also some confusion as to the project's long term goals (other than 'clean portable code'). Daemon news (http://www.daemonnews.org/) receives few articles from NetBSD users - this is really just another symptom of the above. We probably want to concentrate on getting the project information current, though all NetBSD developers and users are encouraged to contribute to Daemon news. We have a list of drivers and the generic devices they support, but few named vendor products as found on the shelves of computer stores. We should try to expand the supported devices list to include more real product names. This could include motherboard/device combinations on the i386 port. There is work underway to make libc threadsafe, and we finally have an open 'tech-smp' list for those working on SMP support. We already spin up secondary CPUs on sparc, and have some code on alpha and i386, but do not have any scheduling machinery in place. All of the free unixes fall short on real regression testing and certification for releases. Independent certification costs significant money, and there are no freely available test suites that cover everything. We have a 'regress' area of the tree - that tends to contain tests that developers have written to assist them in determining when they have fixed problems, but it would probably make sense to see if any of the freely available test suites can provide some verification. It would also be a good idea to run sysutils/crashme against systems before release. Darren Reed darrenr@reed.wattle.id.au is working on a suite for NetBSD to ensure that system calls work in a manner that matches documentation, with both positive and negative tests. NetBSD is one of the harder unixes for a new administrator (though its lack of 'pretty' config tools makes it easier for users to understand what is really happening). We need to expand the documentation for new users. (Volunteers welcomed). This would be welcomed by many users, but it is very difficult to get documentation from some vendors. I believe a team in Japan has made significant progress on a driver. It would be good to determine which are the more popular ports, and for what people are using NetBSD. One suggestion was a 'send-registration' option after install. If the user elected to fill in the form it could also collect the 'dmesg' output, and email to an address @NetBSD.org. Variations include optionally pasting the output into a web form, or emailing the message to a local address for later forwarding. This is a perennial problem - frequently raised and agreed upon, but noone ever seems to want to do the work. Some people are working on to synchronising some kernel interfaces between FreeBSD and NetBSD to make sharing drivers easier. We need to make the use of &man.send-pr.1; more obvious to users (possibly in the INSTALL doc), and also need to work on more timely response to PRs. Interest expressed in an HPPA port. In particular to Erik Berls, Roland Dowdeswell, Charles Hannum, and Alan Horn who manned (and in many cases provided machines for) the booth, and the others who provided machines, time, and effort to make it all work. The NetBSD beer glasses went well. We didn't sell any CDs, which was pretty much expected as every attendee received a free set of CDs courtesy of usenix. We sold 40 out of the 50 official t-shirts - really quite good given the very simple design. imac (macppc) Virtually everyone found this cute. (Many thanks to Charles for working late into the night before fixing an install issue with the new model). Someone from Apple commented that we both booted and shipped unix (NetBSD) on the iMac before they did (Rhapsody). shark (arm32) The digital reference 'Network Computer' design. Possibly even cuter than the imac. Many people asked if they could buy one. uVax3600 (vax) The size of a small fridge (and probably the result of my rental car falling apart on the 1 from La to Monteray), this induced many comments from people, generally about fond (and not so fond) memories, though the number of vaxes still in use in northern europe could be surprising (or not, given the average radiated heat from a VAX). Many thanks to Brian Chase for the loan of this monster. others Questions on just about all ports, including alpha, mac68k, sun3, sparc64, pmax, and hppa (for which we do not currently have a port). The exhibition booth next to ours was Telenet Systems, who among other products produce a range of rack mount i386 PC servers, from a 1U (1.75" high) celeron system for $1099. They expressed interest in having some of their products officially 'NetBSD certified', and also lending equipment to the project for future exhibitions. Charles is following up.