[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

[RFC] Abandoning lowmem level 1

I've just tested memory requirements for i386 and amd64 and updated their 
lowmem levels (amd64 now has separate lowmem levels [1]).

While doing the tests I noticed that there is now only an extremely small 
difference between the limits for lowmem levels 1 and 2 (4MB for both 
architectures), which makes me wonder if we should not just abandon one 

The main difference between the lowmem 1 and full operation was the memory 
required for all translations, but because of the automatic translation 
stripping that is done during anna, the "extra" is almost nil.

Here is the size of the templates file after anna:
- 160110: only English (lowmem 1)
- 277220: English + Dutch (nl)
- 369554: English + Chinese (zh_CN _and_ zh_TW)
- 394446: English + Portuguese (pt _and_ pt_BR)

So, a language adds about 110kB, but apparently that's just enough (also 
given the way cdebconf uses memory) to push the requirement up by 4MB.

IMO a difference of 4 MB is not really worth the testing effort, nor does it 
make any real difference to users. So I propose to just get rid of the 
distinction between lowmem level 1 and 2.


[1] In practice I doubt it matters very much as I'd expect any 64 bits 
system to have more memory than our minimum for normal D-I functioning.

Attachment: signature.asc
Description: This is a digitally signed message part.

Reply to: