On Mon, 2016-09-12 at 19:24 -0500, R. W. Rodolico wrote:
On 09/12/2016 03:53 PM, Michael Tremer wrote:
Hi,
On Fri, 2016-09-09 at 22:54 -0500, R. W. Rodolico wrote:
Actually, I have 10+ systems running on Soekris machines, i586 32 bit AMD Geode's. See http://soekris.com/products/net5501.html. These are not end of life; they are currently being marketed for vpn firewall/routers. I have some that have been running
The form factor of that hardware is nice, but it is really outdated. You can get something way more powerful for about the same money I think.
I just try to build firewalls that last a really long time and use minimum power. However, the same company has a different model (6501 series) that has an atom in it instead of a geode, so I'll just make sure to use those in the future. Slightly higher power consumption (8 watts vs 6 for the 5501), but still not bad.
Modern systems don't use more. Probably even less.
But I think this board is terribly overpriced. It is nice. But I would not even buy it for a quarter of that price shown on their website.
This is a different topic though...
I'm assuming they will not be able to be upgraded after a certain point?
I am not sure if this was clear or not, but this change will only affect IPFire 3. So with that there will be no i686 version.
We will keep running the i586 version of IPFire 2 until we end support for IPFire 2 as a whole.
But we think that the hardware (mainly memory) requirements for IPFire 3 will exceed this hardware any ways and as there are many alternatives available you should not buy anything too old any more. Anything that is about 8 years or newer will probably support 64 bit.
Haven't sold many firewalls recently, so I can live with the restriction on IPFire 3. Yes, you guys are putting tons of work into it, and I can see where it may need more resources on a box running it. I run IPFire 2 on 512M of RAM, and have never run out of memory (but, came within 5% one time during an update). All the things you've been talking about with 3 will require more resources.
It is not strictly needed. I think that IPFire 3 still has a relatively small footprint. But if you look at how cheap memory is there is no reason to save on that. So everyone is "wasting" a little bit. Kernel, user-space, ... and that just adds up. Nobody is restricting themselves any more.
And of course a modern IPFire box has to do way more things than it used to do.
My advice would be: don't buy anything below 2G of memory, without a SSD or with a too small CPU.
That is sad, because these are very stable systems and run nicely with very little memory and extremely low power.
One of the offices I use it at has 50+ employees, 40+ Mbit data line, and some firewall rules built in, along with a site-to-site VPN and a couple or Road Warriors.
All of these systems have pakinfo running on them, so you should see them in the database. Fireinfo appears to show only 13% of the firewalls out there are 64 bit? Am I reading that wrong?
Fireinfo data shows the following (as of today):
- 13% of the systems known to fireinfo are running IPFire 2 x86_64
(http://fireinfo.ipfire.org/statistics/processors)
- 77% of all systems known to fireinfo would be able to run x86_64
(http://fireinfo.ipfire.org/statistics/processors/x86)
So that is a huge majority. I assume that many of the VMs are just running as 32 bit and probably would support 64 bit, too. So that number could be close to 90%.
I'm guilty of that; I have some vm's running which are quite possibly 32bit, but they are on 64 bit machines and could likely be upgraded easily. Just add RAM :)
But, I do see where ARM is not happening.
In fireinfo ARM basically does not exist. Sadly.
Rod
On 09/08/2016 12:35 PM, Michael Tremer wrote:
Hi,
On Thu, 2016-09-08 at 16:30 +0000, IT Superhack wrote:
Hello Michael, hello development-list,
I have a question concerning the IDS2016 (URL: http://wiki.ipfire.org/ids/2016 /log), where it says: "Dropping i686 + armv5tel - No need for it any more"
Yeah, we didn't really log everything what we talked about. There was a little bit more than that and there is probably a little bit more detail to all of it than that.
But generally we didn't make as many decisions as we used to do since we have the monthly telephone conference.
To me, it has not become fully clear if this means the architecture or the whole release format. For example, nearly nobody (Fireinfo says: 0,07%) is running an IPFire system with the armv5tel architecture - these are afaik mostly old systems like the Raspberry Pi which are certainly not suitable for a firewall purpose.
So to give a little bit more detail:
The plan is to drop all 32 bit architectures as soon as possible. We do not see any point in supporting these any longer.
That starts with ARM where we never really got a number of users that is high enough to justify all the effort that is going into development of this.
And secondly x86: All hardware that is bought today or in the last ~5 years will support 64 bit. If it doesn't and if someone bought an other ALIX device that is just bad luck. These are too slow to run an IPFire system decently any ways and there is better alternatives on the market.
That is why we do not see any point whatsoever to continue supporting these architectures.
We will keep armv7hl for now because there is a some hardware around and our build system use it, but this will probably go away very soon when there is no usable hardware around soon.
On the other hand, > 80% run an i686 system, and I guess it wouldn't make sense to make these installations unusable because of EOL. But maybe "dropping" means that you will remove some specific patches for i686, so these will run with i586 afterwards.
No, this will scrap support for i686 entirely.
Within the ARM stuff, the situation is not that clear for me. Are you panning to remove the ARM support at all? Or are you going to remove ARMv5 devices from the "supported ARM devices list"?
We actually have done that in the past and we did not add any new hardware support in the last few years.
There is neither support nor any requests from the community for this.
Feel free to leave your comments on this. Would be happy to hear if someone can come up with at least one argument to continue 32 bit support. We couldn't find a single one.
Sorry if there is a misunderstanding here - got way little coffe today. ;-)
Best regards, Timmothy Wilson
Best, -Michael