News for hackers. Stuff that matters.
Re:A faster BBC News front page by acdha on Monday April 24, 2017 @03:45PM (#14187450)
> Most of the work would be essentially quick and one-off:

Again, think about the complexity of what's generating that content. Do you really think they have a single index.html template floating around rather than the system described pulling blocks of content in from various sources, all of which need to be updated in various places? You could build a system to collapse whitespace runs after the content is generated but now you need to develop that and test it to make sure it doesn't trigger on something where the whitespace actually matters … and suddenly you're taking time away from changes which actually benefit people.

> given an almost certain quick small win + saving CPU time and bandwidth and a bit of planet while we're at it!

How much CPU time do you think you're talking about, anyway? Here's what doing this in a naive manner which will break things looks like:

    cadams@jupiter:~ $ gzip -9 < bbc.html | wc -c
       32259
    cadams@jupiter:~ $ curl -s http://www.bbc.com | perl -pe 's/\s+/ /gs' | gzip -9 | wc -c
   30859
So we're talking an extremely small amount of CPU time and one packet saved. That's might seem useful until you look at this in context of a page which loads a couple MB of content and uses multiple seconds of CPU time:

https://www.webpagetest.org/result/170424_NT_W4S/

Image compression, enabling gzip on the remaining resources which lack it, etc. are all going to save more data on individual files and have the nice property of not breaking anything else.

Re:A faster BBC News front page by DamonHD on Monday April 24, 2017 @04:51PM (#14188188)
I'm not saying don't do the other things too, but a ~5% saving (maybe more at a more typical gzip TX setting) for no semantic change at all saves some outgoing BBC resource but is more significant for the RX side in CPU, etc.

Anyhow we can agree to disagree but many of my clients over the years have been very happy with 5% resource savings, although 5x is nicer when available... B^>

(And yes I'm fudging CPU and bandwidth, but with a compressor involved and the CPU often being a bottleneck on lower-end devices for example, I think that simplification can be justified.)