1 |
On 08/01/2014 04:52, Raúl Porcel wrote: |
2 |
> On 07/26/14 19:33, Michael Palimaka wrote: |
3 |
>> On 07/27/2014 03:19 AM, William Hubbs wrote: |
4 |
>>> If an arch team isn't going to honor a stable request, shouldn't they |
5 |
>>> remove themselves from it and say so? |
6 |
>>> |
7 |
>>> Also, if an arch team does that, does that mean we don't have to file |
8 |
>>> stable requests for that arch on future versions of the package? |
9 |
>> |
10 |
>> When armin did stabilisation for minor archs in the past, he took the |
11 |
>> opportunity to evaluate whether it was still useful to have the package |
12 |
>> stable. In many cases for small random packages, stable keywords were |
13 |
>> dropped to reduce future workload. I always thought it was a pretty good |
14 |
>> strategy. |
15 |
> |
16 |
> Indeed! The thing was that a lot of the packages were keyworded and |
17 |
> marked stable back in the day where the arch was more popular. |
18 |
> |
19 |
> But almost all arches except amd64/x86/arm are getting less and less |
20 |
> popular: |
21 |
> |
22 |
> alpha: no new hardware in more than 8+ years |
23 |
> hppa: being phased out IIRC, and no new workstations(ie, graphics/sound) |
24 |
> in 5+ years |
25 |
> ia64: no new workstations in 10 years, new servers are expensive |
26 |
> ppc*: new workstations are expensive |
27 |
> sparc: no new workstations in 7+ years, new servers expensive |
28 |
|
29 |
Who says they have to be new? Sometimes, the fun is in the old hardware of |
30 |
yore. I found out last week that sparc32 is still quasi-alive, though it |
31 |
doesn't appear to have any mainstream kernel maintainers. |
32 |
|
33 |
ia64, go search on eBay for old SGI Altix/Prism gear. There's a metric ton |
34 |
of Altix units being offloaded lately. There was one listing a few weeks |
35 |
ago that had 10-20 Altix 350 servers, dual or quad CPU, for ~$90 per server. |
36 |
Even saw an SGI Prism a few days ago (which is just an ia64 variant of the |
37 |
Tezro). |
38 |
|
39 |
|
40 |
> One of the reasons they are being killed, IMHO, its that the power |
41 |
> consumption isn't worth, and an amd64 machine is pretty much more |
42 |
> powerful, has more cores, and cheaper and has a lot less power consumption. |
43 |
|
44 |
This is always a concern. I used to run equipment 24/7, but that took |
45 |
chunks out of my electric bill each month. Now, I sleep both of my Intel |
46 |
systems and power down the SGI machines when they're not in use. I |
47 |
eventually need to test out hibernation on the SGIs and see if that can be |
48 |
of any use. Ctrl+Z in the middle of a compile, then hibernate...in theory, |
49 |
I should be able to resume and then 'fg' the compile back into action. |
50 |
|
51 |
|
52 |
> My Sun Blade 1000 (workstation) uses 225W idling, my amd64 workstation |
53 |
> uses 100W at full power or so. And the amd64 has way more cores and more |
54 |
> performance. And let's not talk about the heat... |
55 |
|
56 |
SGI O2, 1x CPU, 512MB RAM, 2x HDDs - ~80W |
57 |
SGI Octane, 1x CPU, 2GB RAM, 3x HDDs - ~303-330W |
58 |
SGI Onyx2, 4x CPUs, 8GB RAM, 5 HDDs - ~720W |
59 |
|
60 |
I ran some very quick calculations on running those three systems full time, |
61 |
24/7 for a month, along with my two Intel Linux systems (~160W and ~140W), |
62 |
about ~$0.10/kWh (distribution charge only, did not factor in transmission |
63 |
rates nor taxes), and it should only add an extra ~$120 to my bill per |
64 |
month, which isn't that bad (but then again, I am serviced by a co-op that |
65 |
has low rates). So running them less often should be easily manageable. |
66 |
Probably moreso in winter, as rates are bit lower then (gas isn't, though, |
67 |
but SGIs make great space heaters). |
68 |
|
69 |
|
70 |
> Besides there's software like firefox and gnome3 that doesn't work in |
71 |
> sparc due to unaligned accesses. |
72 |
|
73 |
Pft, if people want those shinies, run them on a standard PC. If the only |
74 |
point of having an alternate arch is so you can surf the web...then I don't |
75 |
see much of a point. The direction that most of the large projects are |
76 |
going in is not conducive to old equipment anyways. There is, however, |
77 |
always going to be the smaller projects that might be tickled pink on having |
78 |
their software run on old hardware. Like maybe someone taking fvwm and |
79 |
tailoring it to look like IRIX's 4dwm. Something I've thought about if I |
80 |
ever get bored enough, and can fix all the other bugs I keep running into on |
81 |
these systems. |
82 |
|
83 |
|
84 |
> Debian announced some months ago that they're dropping sparc support as |
85 |
> well. Right now debian doesn't support, officially, alpha, hppa and sparc. |
86 |
|
87 |
Their loss :) |
88 |
|
89 |
-- |
90 |
Joshua Kinard |
91 |
Gentoo/MIPS |
92 |
kumba@g.o |
93 |
4096R/D25D95E3 2011-03-28 |
94 |
|
95 |
"The past tempts us, the present confuses us, the future frightens us. And |
96 |
our lives slip away, moment by moment, lost in that vast, terrible in-between." |
97 |
|
98 |
--Emperor Turhan, Centauri Republic |