Dani M.'s picture

I've seen some past discussions on this in the blog (or the forum)...

I would like to "vote" against this deprecation. Or better, argument against it.

In summary, I've read that dev team does not really see a good reason for mantaining the 32 bit options given today's professional hardware... And they are taking lots of hosting room (lots of appliances and  flavors for each of them).

The 32 bit options have been "hidden" from the website (need to Google for reaching them). It is announced that they are deprecated, and they will be surely removed on next release... unless community asks enough for maintaining them.

But I think that Turnkey Linux should fit the user with minimum requirements: on knowledge, and/or possibly also on hardware. The user has not always chosen his lack of knowledge, but nor his lack of hardware/ressources.

I think that the TLKX should assume the less possible on the user, trying to make straighforward to use it in the wider number of reasonable cases (or at least the most basic ones). It's really about "democratizing" something not trivial for the masses. If it covers more specialized professional cases, this is great, but this shouldn't be a detriment to the "base case" (minimal assumptions).

As an example, I'm a teacher from Spain. We are "specially" in crisis lately, but I still consider my lucky country part of the "upper-middle world class". Tons of countries and regions in the world are even much much worse.

And given my experience as a teacher in my country, I know that a lot of classroom PC's (desktops) are still 32 bit hardware. 

Or they may have 64bit hardware, but still have with 32 bit OS on them.

Or they may have 64bit hardware and 64bit OS, but with Hardware virtualization that cannot be enabled on the BIOS, so Virtualbox sadly does not permit to install 64 guests in that case.

And for sure, the same occurs in many "students" home PC's, and I assume that it happens on many home Desktops of possible users interested in TKLX... And it happens on lots of home or SOHO "old hardware" that could have a (really easy) new life with a TKLX .ISO...

So having currently an enviroment at school suitable for the currently pushed 64bit version is not at all evident in my experience (and foreseable futur). And it cannot either be assume universally for the homes.

These are future IT guys that will have for sure some 64 bit hardware at work (I hope so)... But they are learning today with what they have. Or anyone may have "at hand" only a 32 bit possibility to do some initial testing.

These are users that may not hang a lot in the forums (some novice profiles, and not so novice, don't "talk" very much)... but I guess that there may be a considerable amount of "hidden" and silent users like this.

For sure the 64bit was really needed (and requested), for the "pro" side of TKLX users... But I think that at least some of the light 32 bit option should be left also... for the weakest, but also for everyone's freedom.

As a teacher, I'm thinking in comparison on the really bad/blind move that Microsoft made on Windows Server 2008 R2, by deprecating 32 bit support on that release and later ones... This makes currently quite difficult to work with Windows Server in the classroom (or assume that all the students can do the stuff as homework). Teachers would naturally stick to the light 2003 (discarding Vista based 2008). What can I do now that 2003 has reached end-of-life? The school has no money for upgrading hardware, some students neither... Should I teach deprecated Windows Server OS (2003/2008)? Or directly forget about Windows Servers, and just focus only on Linux... Bad for Microsoft, good for Linux... and in principle for TKLX :-)

But it becomes increasily difficult for me and my students to use TKLX, and worst if 32bit is finally removed. It would really be a pity. Well, for sure I can now use tkldev to prepare the VM's to the students (not dived yet on that)... but I prefer to show them things that they can do exactly themselves when they are out... And using tkldev is not so "turnkey", for sure (at least in comparison with using directly the built stuff, as currently is possible).

If Turnkey Linux becomes less turnkey, the sadly normal move is towards other options: much horrible "turnkey" alternatives (Bitnami and the like, that still support 32 bit), or alternatives with stepper learning curve, but that do not albitrarily "damn" my "obliged" 32 bit usage: a "standard" Debian from scratch, or a Vagrant/Puppet (or the like).

Being (silently) with TKLX for these recent years, I would see this really sad (because I think that TKLX is really a great project...).

Microsoft can do what he likes, even if they are wrong assuming that using an OS in production is the same as learning it. But Linux is more than that: it has 0 price, 0 assumptions, so it makes sense using it even "on production" on cheap hardware, if it suits someone's needs. A server is not always a powerful machine, it is just a specialized one (see the Raspberry Pi and the like uses for clear examples). It is great to reuse old hardware, and doing it "easily" (there is some ecology here also).

For these reasons, I would ask for TKLX not abandoning the 32 bit generation for "easy usage".

But I would go further: currently the 32 bit are really hidden (not novice-friendly at all)...

Maybe this has made TKLX loose some newcomers since the change to 64 bit: a user tries the propoposed download, it does not work (not for his fault, but because he has an old hw or OS, he may even not know)... He may (silently) go to search elsewhere. So the pushed 64 bit change may have won some new users (avid 64 bit claimers, that correctly argumented that not having 64bit was bad for Turnkey Linux popularity).... but it may also have lost some other ones (and it was really not needed).

I would change this and in fact "re-ad" the links to the 32 bit builds (or some of them) on the Turnkey Website, and recommend them for the novice users. They will be surely the builds giving less problems to them (to start learning): they will work on any (novice) virtual machine, and easily on every bare metal hardware they have spare (I don't think that the UEFI hardware with no BIOS emulation enabled is going to be a "spare bare metal" machine for these users). 32 bit is currently the most universally working environment. It is what less assumptions makes (at least for an initial use on a testbed).

I completely understand the issue with hosting space and appliances multiplication (this is crazy)...

But I think that at least some "key" 32 TKLX builds should be mantained. The ones more universal and the ones more geared to novice users. From my point of view, these are:

* .ISO (covers even bare metal, and all of the VM "SOHO" solutions).

* The current .OVF zip (currently the most "straighforward" TKLX usage, by using Virtualbox... I've talked about this on a different thread).

If only one of both builds can survive, I would (sadly) choose the .ISO one (covers bare metal)... but with only this, one loses the currently most straighforward use (VBOX VM from OVF). Really a pity!

If even these two 32 bit builds are too much for all the appliances, I would consider at least still provide them for the more basic or popular appliances. Or sacrifice/merge other builds (I've talked in another thread about the currently apparent redundance of the two ZIP builds, or maybe there are some of the virtualization platforms that are indeed less potentially used than the 32 bit "basic ones").

Just some feedback for the core TKLX team, from an "external" point of view. Hope that it helps the project. :-)

 

Forum: 
Jeremy Davis's picture

Although TBH that has become less of an issue more recently as we now have a pretty good mirror network. Initially we relied on SourceForge to be our primary mirror and they were constantly asking us to keep the size of the library down.

And although there is some payoff to supporting less architectures (less maintenance overhead), really it doesn't make that much difference IMO (most of it is automated anyway...).

So whilst technically we have decided to drop 32 bit support, there has been no final decision on the timeframe of that. Initially the plan was to drop it for the v13.0 release - but upon reflection and discussion we decided to leave it be for now. So the idea was to drop it for v14.0 release, however that's not set in stone either...

Again if you want to post something brief on the Issue tracker that would be appreciated. thanks for your input! :)

Dani M.'s picture

Specially because I still see the need for it.

Besides, if 32 bit links are not reintroduced in the App's download pages (or while they are), I would strongly recommend a brief note on these pages stating that downloads are being 64 bits. It could be just a "(64bit)" string, after the word "BUILDS".

Currently it is not clearly said (only seen when saving the file name, if one is tech-savvy and "decrypts" the file name). Even if TKLX are relatively small files (for being appliances), it is not nice to download something just to discover that it does not work (and that possibly one cannot do anything to solve it, since it depends on one's own HW/OS).

I will also put a little note on this on the "docs/build" page (at least reflecting current state).

My personal "traumatic" experience with TKLX 64 last year was that I discovered that you changed to 64bit in "live" in the classroom: just downloaded the last one at home (did not even read the file name... I was just trusting, you know), tried to open it on the classroom PC's in front of the students... And bang! Went to turnkey website, no 32 bit links, no mention of 64 bit... Needed to google to find the 32 bit versions and wait for them (and later found the posts regarding its deprecation).

I imagine lots of my novice students (and lots of users out there), just blocked by this: download, does not work on my computer, abandon...

32 vs 64 awareness is increasingly becoming mainstream. If at least "64 bit" is clearly seen before download, "some" of the users may at least avoid the waste (download, unpacking, possible VBOX installation, and things do not work).

 

L. Arnold's picture

For instance. ESXI on a hp dl380 wont virtualuze 64 bits though it will as a stand alone server. 32Bits is fast enough for many iterations and folks who have latest version servers may not be the ones getting into TKL as the example of Syudeny use so aptly gives. I would expect that TKLBam more than offsets cost of storing 32bit builfs buy perhaps I am wrong there. Keep 32 BIT ot ay least documeny where the last version is or how yo builf yhemm. Mobile typos sorry. Thanks for the system. We appreciate it.
Jeremy Davis's picture

One way to go with your old 32 bit builds might be to do an in-place Debian upgrade?! IIRC v12.x was our first Debian based release and was based on Debian 6/Squeeze. I recently heard that 7/Wheezy is having an extended LTS so active repos should still be available. And 8/Jessie is still in LTS for another few years. So doing a step-by-step dist-upgrade process should be possible. I.e. dist-upgrade to 7/Wheezy, then make sure everything is good (fix anything that isn't), then dist-upgrade to 8/Jessie (fix), then finally dist-upgrade to 9/Stretch (and final fix).

One issue you'll likely hit is that we haven't built all our v15.x/Debian 9/Stretch packages to 32 bit. But 14.x/Debian 8/Jessie will be supported by Debian for another few years yet, which might be enough for your purposes? Also if it's really old hardware, note that Debian (as of 9/Stretch) now only supports the 686 (or later) CPU instruction set in their 32 bit builds. That means that really (really) old 32-bit hardware (i.e. 586 or earlier) simply won't work (almost certainly won't boot).

Regardless, if you go that path, I urge you to make a backup before you start. If these are installed to bare metal (which seems to be the case), then I suggest Clonezilla or similar to take an image of all the data. Furthermore, make sure you test it (e.g. in a VM).

If you're happy to do data migration, then you'll find most of the library as v14.1 ISOs on the mirror. They were only ever built as ISOs and there are a few missing IIRC, but LAMP is definitely there. If you're browsing, then just look for 14.1-jessie-i386.

Another option would be to download the most recent (v14.1) 32-bit TKLDev (it's here on the mirror) and use that to build an updated appliance. TBH because of v15.0 changes, I doubt it will just work. But if you want to go that way, please download it and have a go. If/when you hit issues, I'll do my best to help you out. If you do that, you could likely get a v15.0 running on 32 bit (and build any required 32 bit packages).

Add new comment