[OpenIndiana-discuss] [developer] Re: SPARC-OpenIndiana Screenshots

[Sašo Kiselkov]

On 09/10/2012 02:49 PM, Martin Bochnig wrote:
> I know, But on everything starting at 1GHz it is not that bad.
> The reason was: Of course I'm aware of the smaller compression ratios.
> But then DualLayer would be a requirement.
> If you like and if time allows, I can offer two versions of that ISO.
> One for DualLayer Media (with gzip-5) and the other for usual 4.7 GB
> (4.4GB) media.

Contrary to popular belief, the -[1-9] switch to gzip doesn't influence
compression ratio all that much. Here are my tests with the silesia corpus:

$ time gzip -1 -c < silesia.tar > silesia-1.tar.gz
	user	0m43.004s
	ratio	63.5%

$ time gzip -3 -c < silesia.tar > silesia-3.tar.gz
	user	0m52.029s
	ratio	65.5%

$ time gzip -5 -c < silesia.tar > silesia-5.tar.gz
	user	1m9.355s
	ratio	67.4%

$ time gzip -6 -c < silesia.tar > silesia-6.tar.gz
	user	1m29.819s
	ratio	67.9%

$ time gzip -9 -c < silesia.tar > silesia-9.tar.gz
	user	3m16.998s
	ratio	68.1%

I hardly think the 2.5% difference in compression ratio is going to push
you over the single-layer limit. By a first-degree approximation, 10GB
will roughly compress like this:

	gzip -1: 3.64 GB
	gzip -3: 3.45 GB
	gzip -5: 3.26 GB
	gzip -6: 3.21 GB
	gzip -9: 3.19 GB

So save yourself a ton of time and use something between gzip -3 or gzip
-5. Anything higher than that is placebo.

Cheers,
--
Saso