Okay, I'm going to ask the gajillion-dollar question... I've been doing syslinux for 11 years now. From a very limited scope it has since grown and is now a very advanced boot loader. Unfortunately, it's not clear to me anymore to what extent the work I'm doing is useful. I personally dislike grub because of its monolithic design, and because the grub people have traditionally been very difficult to deal with. Yet it has achieved a remarkable adoption rate, and it's not clear whether or not it's worth trying to capture that market (especially since these are markets in which you don't get paid...!) In many ways Grub is like Windows; bloated, bad design, has the checkbox features, most of them badly implemented, and enough pretty pictures people think it's user friendly. I don't want to turn syslinux into that. So I guess the megaquestion is... Is it worth bothering with going forward, or am I wasting my time? If so, what should I be focusing on? -hpa
On Fri, 7 Jan 2005, H. Peter Anvin wrote:> Unfortunately, it's not clear to me anymore to what extent the work I'm doing > is useful.<snip>> Is it worth bothering with going forward, or am I wasting my time? If so, > what should I be focusing on?I guess the real niche I see your efforts filling (admirably) is for small, approachable loaders, suitable particularly for installers, and slender implementations. I note the present time is a bit after 5 am EDT, and 2 am PDT -- sounds like a lonely 2 am thought to me. I wish I could call to say: Thanks !! I appreciate your work; I use your work every day, and I would miss it greatly if it became unusable as hardware evolved. Its presence and quality (with busybox) makes small, niche loaders and distribution creation possible. It is clear that large parts of the future will move to smaller (possibly wired, possibly IR or wireless or bluetooth [and whatever the next such technology will be] connected) smart devices, probably acting as clients of fewer central device(s) -- this client/server model is the only one providing for reasonable and scalable deployment, and management. The deployment of RFC 1918 and self-organizing Zeroconf networks, and the ongoing IPv6 infrastruture buildout, combined with FOSS implementation from the bottom of the stack, and continuing up into the meta-app layer have made this doable and probable. Seeing this prospect, I build my parts toward, and (eating my own dogfood, during development) have 'lived in' parts of that future for many years, with the Syslinux tools a part of that existence. Could I transition to grub? -- maybe -- but I took a run at the topic, 12-18 months ago to expand my coverage [and not to replace or displace Syslinux based tools], and found the documentation opague, incomplete, stale, and incomplete. Yuck. That said, you have been developing in a reactive mode, rather than from plan in the last year or so. This has to be tiring. If you are not feeling 'the itch', it may be profitable to consider what future you want to attain. Anyway, I appreciate your work. -- Russ Herrold
Hi H. Peter,> -----Original Message----- > From: H. Peter Anvin [mailto:hpa at zytor.com] > Sent: Friday, January 07, 2005 9:50 AM > To: SYSLINUX at zytor.com > Subject: [syslinux] syslinux vs grub > > Okay, I'm going to ask the gajillion-dollar question... > > I've been doing syslinux for 11 years now. From a very > limited scope it > has since grown and is now a very advanced boot loader. > > Unfortunately, it's not clear to me anymore to what extent > the work I'm > doing is useful. > > I personally dislike grub because of its monolithic design, > and because > the grub people have traditionally been very difficult to deal with. > Yet it has achieved a remarkable adoption rate, and it's not clear > whether or not it's worth trying to capture that market (especially > since these are markets in which you don't get paid...!) > > In many ways Grub is like Windows; bloated, bad design, has > the checkbox > features, most of them badly implemented, and enough pretty pictures > people think it's user friendly. I don't want to turn syslinux into > that. So I guess the megaquestion is... > > Is it worth bothering with going forward, or am I wasting my > time? If > so, what should I be focusing on? > > -hpaI understand your comments all too well. We at leaf.sf.net, Bering-uClibc router software, do use syslinux and isolinux to run our software and occasionally pxelinux to startup some embedded router hardware. I find your work invaluable and appreciate this 'Linux habit' you have. Let me finish by saying that for embedded hardware and for rescue disks, syslinux will continue to live on! Don't despair, keep on! Luis Correia Bering uClibc Team Member PGP Fingerprint: BC44 D7DA 5A17 F92A CA21 9ABE DFF0 3540 2322 21F6 Key Server: http://pgp.mit.edu
On Friday 07 January 2005 10:50, H. Peter Anvin wrote:> Okay, I'm going to ask the gajillion-dollar question... > > [...] > > Is it worth bothering with going forward, or am I wasting my time? If > so, what should I be focusing on? > > -hpaI personally dislike grub because it's difficult to set up (IMHO)... I've tried to set up GRUB half a dozen times and always failed, only recently I've managed to get it working... The things I like about SYSLINUX is: - it's easy to set up - the configuration (file) can easily be changed without the need to re-run the installer (unlike LILO, for example) - the configuration is easy to write in the first place (unlike GRUB) - It Just Works (tm), never had any problems with it - it's extensible through COM32 programs (although we've never needed that) Let me explain why and how our company uses SYSLINUX (note that I'm actually working for a company called Be OK/Comdasys, I only use the POP account of our partner Webport for mailing lists ;-) We're using SYSLINUX to boot our products (VPN gateways with VoIP proxy). One thing we're doing is having a four partition layout: one boot partition (FAT16, SYSLINUX), two main partitions and one /etc paritition (all basic packages like sysvinit have been patched to not use /etc but another directory which lives on the main partition so the system can boot without an /etc directory). Now one main partition is in use and the second one is where updates are installed into. Once an update was installed successfully into the unused main partition my installer mounts the FAT16 partition, sed's the syslinux.cfg and reboots. This way we ensure that the system is always bootable, even when the update failed (in that case the syslinux.cfg won't be touched and the old main partition is booted). The only critical point is when the syslinux.cfg is rewritten, the system may crash at any other point in the update process and it is guaranteed to boot up fine. I don't know a way to do something like this with any other boot loader. C'ya, Marc -- Marc Haisenko Systemspezialist Webport IT-Services GmbH mailto: haisenko at webport.de
Hi Peter,> > Is it worth bothering with going forward, or am I wasting my time? If > so, what should I be focusing on? >I'm certainly not a guru, nor have I any specific computer science knowledge (I'm just a chemist). However, I've been using syslinux (isolinux) to make boot CD's for my linux boxes since quite a bit. From my really *narrow* viewpoint: - it works absolutely great - it's small - the first time I've had an installation problem, you were so prompt in helping So, Peter, focus on whatever you want, but, please, keep on maintaining this great piece of software!!! Thanks and ciao -- Gabriele Balducci - Dipartimento di Scienze Chimiche - Via L. Giorgieri 1 I-34127 TRIESTE - tel: I-040-5583957 - fax: I-040-5583903 e-mail: balducci at univ.trieste.it
On Fri, 07 Jan 2005 01:50:26 -0800 "H. Peter Anvin" <hpa at zytor.com> wrote:> Okay, I'm going to ask the gajillion-dollar question... > > I've been doing syslinux for 11 years now. From a very limited scope > it has since grown and is now a very advanced boot loader. > > Unfortunately, it's not clear to me anymore to what extent the work > I'm doing is useful. > > I personally dislike grub because of its monolithic design, and > because the grub people have traditionally been very difficult to deal > with. Yet it has achieved a remarkable adoption rate, and it's not > clear whether or not it's worth trying to capture that market > (especially since these are markets in which you don't get paid...!) > > In many ways Grub is like Windows; bloated, bad design, has the > checkbox features, most of them badly implemented, and enough pretty > pictures people think it's user friendly. I don't want to turn > syslinux into that. So I guess the megaquestion is... > > Is it worth bothering with going forward, or am I wasting my time? If > > so, what should I be focusing on?I hadn't noticed syslinux until I started trying to make my usb memory stick bootable. The thing that made it more attractive to me than grub though was that it's config syntax was so similar to lilo. I'm like you, in thinking that Grub is kind of bloated and tries to hold your hand too much. Also, the new Debian-Installer and Knoppix (and FeatherLinux, which is based on Knoppix) all use syslinux for booting from usb memory stick, cdrom and other boot media. I find syslinux to be a great medium between "ultra basic, barely does what you want" and "bloated, hard to use, but when you get it working, it has more gui and clicky-things than Windows". It has the nice boot splash and allows you to specify help files for the user to read at boot time. I couldn't be happier. :-) So, thanks for all your hard work! Jacob
> I personally dislike grub because of its monolithic design, and because > the grub people have traditionally been very difficult to deal with. Yet > it has achieved a remarkable adoption rate, and it's not clear whether > or not it's worth trying to capture that market (especially since these > are markets in which you don't get paid...!) > > In many ways Grub is like Windows; bloated, bad design, has the checkbox > features, most of them badly implemented, and enough pretty pictures > people think it's user friendly. I don't want to turn syslinux into > that. So I guess the megaquestion is... > > Is it worth bothering with going forward, or am I wasting my time? If > so, what should I be focusing on?I'm grateful to you for working on syslinux and especially on PXELinux. Thanks to your latest improvements (simple menu system, multiple TFTP servers, NOESCAPE flag) PXELinux has all the functionality of PXEgrub/nbgrub and the configuration/installation is much simpler. Ethersel (although you wrote it for syslinux) provides even more functionality: Using it with PXELinux and etherboot .zlilo images (with modified VCIs) enables me to build a boot menu for all PCI NICs in my network where I can choose between different DHCP servers (!) without having to deal with a list of MAC addresses or other complicated stuff. PXELinux has the *very* big advantage, that is uses the UNDI stack (is that the proper term?) of PXE/Etherboot and does not need to be recompiled when new a type of NIC is added to network. Grub on the other hand has to be compiled with specific NIC support and becomes unstable, when support for a lot of NICs is compiled in. Grub support for new NICs is only available via 3rd party patches and they are not integrated into Grub. Even if the Grub maintainer would accept the patches, PXEGrub's support for NICs would be limited to the ones supported by Etherboot (grub uses the etherboot drivers). PXELinux on the other hand does support every NIC that has a PXE rom and therefore will support every *upcoming* NIC with such a rom. So when it comes to network booting PXELinux rules! Alex
Peter, Thank you very much for all your hard work. PXELinux is an integral part of our validation test network, it is the only software that does what it does and actually works on all of our hardware. We take full advantage of PXELinux, memdisk, and Muralis(thanks to you too Murali) menu system. I cant wait to try the new menu system and all of the new features you've been creating. You're not wasting your time. I think you should focus on taking a break, the code can wait. -Aaron On Fri, 7 Jan 2005, H. Peter Anvin wrote:> Okay, I'm going to ask the gajillion-dollar question... > > I've been doing syslinux for 11 years now. From a very limited scope it has > since grown and is now a very advanced boot loader. > > Unfortunately, it's not clear to me anymore to what extent the work I'm doing > is useful. > > I personally dislike grub because of its monolithic design, and because the > grub people have traditionally been very difficult to deal with. Yet it has > achieved a remarkable adoption rate, and it's not clear whether or not it's > worth trying to capture that market (especially since these are markets in > which you don't get paid...!) > > In many ways Grub is like Windows; bloated, bad design, has the checkbox > features, most of them badly implemented, and enough pretty pictures people > think it's user friendly. I don't want to turn syslinux into that. So I > guess the megaquestion is... > > Is it worth bothering with going forward, or am I wasting my time? If so, > what should I be focusing on? > > -hpa > > _______________________________________________ > SYSLINUX mailing list > Submissions to SYSLINUX at zytor.com > Unsubscribe or set options at: > http://www.zytor.com/mailman/listinfo/syslinux > Please do not send private replies to mailing list traffic. > >
"H. Peter Anvin" <hpa at zytor.com> writes:> Is it worth bothering with going forward, or am I wasting my time? > If so, what should I be focusing on?For installation and rescue boot disks, SYSLINUX is already the tool of choice. In my opinion, this is primarily because it runs in real mode using BIOS calls to access all hardware. This lets a generic boot image work no matter what hardware happens to be in the machine. The ability to use a modern network card's UNDI stack during PXE boot is partcularly important; for this reason, I really had no choice but to use SYSLINUX for Unattended. GRUB's big strength (relative to the boot loader it replaced, LILO) is that it can read files from the hard drive without the need to run an installer. When I accidentally hose my boot loader, a GRUB rescue floppy lets me boot my system again by hand. I even get name completion on file names as I manually tell it which kernel and initrd to load. SYSLINUX can read the file system, too, but until recently, only FAT. You added ext2, which is good. But not all the world is FAT or ext2. Are you going to add support for XFS, JFS, etc.? And even if you do, with the current design each would (I assume) be a separate boot loader... Which means a truly generic boot floppy would remain impossible. So, in my usage scenarios, to supplant GRUB you need two things: 1) A better boot-time command-line interface (e.g., the ability to select the boot device, TAB completion of file names, or at least the ability to browse the filesystem) 2) Modular filesystem drivers (I assume you would eschew a GRUB-like monolithic design). So I could boot a FAT floppy, tell it to load the appropriate ext2/xfs/jfs/whatever module, then load and boot the kernel from my real hard drive. If you managed to start a wave, maybe other people could be convinced to do the other file system modules. Personally, I would love to see GRUB replaced. Anyway, those are my musings. Flame away! - Pat http://unattended.sourceforge.net/