On Sun, Jun 01, 2025 at 07:21:35AM -0700, Ron Minnich wrote: > One thing I hear a lot: installing Plan 9 is headache for many > people, particularly trying to trace down working docs to produce > working code: "... it's just a maze full of dead ends." > > It's like building a sofa from several different IKEA flatpacks > without instructions :-) > > I know lots of you have done lots of work to make this better, but > we're not there. t's easy for many > of us, at this point, because we have the muscle memory. For anyone > new to this, it's just > baffling: their muscle memory is for linux. > > I think one thing we could put up at the p9f.org site is a set of > simple instructions that include: > - using the WASM version in your browser > but I'd like to see this using Lola, not rio. I've lost more people > when I show them rio ... > whether or not it's great, it's too different for many people. The > lessons rio contain are too > much for many people taking a first look. > - setting up with qemu. I'd most prefer to have a git repo which > contains both an ISO and source to > drawterm which *builds*. I continue to get reports of build problems > with drawterm. If there is a > devdraw dependency, put that in the repo too. I mention github > because it's linux friendly and > I can set up CI to ensure that whatever people are pulling, it seems to > work. > - have some way to show people they can "cpu" into the qemu instance, > and show off > the power of that model. > - There are too many fiddly bits to get right to bring up cpu and > other services. I > > The challenge: anybody want to take on any of these things? > > I may try doing the github repo, since most complaints I get revolve > around qemu and linux usage.
On my side, I will continue, progressively, to feed the nix related pages: http://notes.kergis.com/nix-os.html with my own notes while I stumble upon installation problems (because, eventually, Plan9 and Nix have to go back to real hardware). There are indeed discrepancies between the various documents one can find online, and multiboot (one problem when installing on real hardware) can be a headache: 9front has GPT support, but not 9legacy, and with GPT the partitions have to have a different GUID (I don't speak about the type of the partition but of GUID of the partition itself) or all the UEFI booting managers will get confused facing two different partitions with the same GUID (at the present time, an installation is not reentrant: you can't install two Plan9 systems, independent, concurrently on the same node; the installation scripts will be confused; the booting code also). And 9legacy disk/fdisk still messes up an existing partition by recomputing the starting block of partitions it had not created, rendering other partitions not findable (user may think that he even lost data or even other installed systems since the other systems will generally not boot anymore not finding a correct first block to chainload). UEFI[*], for the last booting part, seems to have been adopted by ARM and RISC-V, there are probably scripts that could be written to explore the hardware, send the description to some builder to compile a kernel and allow network booting too. *: I don't specially like UEFI: it is too big, convoluted, with a verbose style that I dislike, and it is not really open: if one tries to contribute, one faces the pillow strategy: just ignoring anything sent, silently killing contributions waiting for the tentative contributor to simply pass his way), UEFI, since for the booting part (there are two parts). -- Thierry Laronde <tlaronde +AT+ kergis +dot+ com> http://www.kergis.com/ http://kertex.kergis.com/ Key fingerprint = 0FF7 E906 FBAF FE95 FD89 250D 52B1 AE95 6006 F40C ------------------------------------------ 9fans: 9fans Permalink: https://9fans.topicbox.com/groups/9fans/Tbe8e5fda6ae62f5c-Mbebc8b890a7aba2bd4592bf2 Delivery options: https://9fans.topicbox.com/groups/9fans/subscription
