deeply impressed with pagespeed (and nginx)

19 views
Skip to first unread message

Luke Kenneth Casson Leighton

unread,
Aug 25, 2017, 7:44:42 AM8/25/17
to mod-pagesp...@googlegroups.com
ok so i've been running mod_pagespeed for a couple of weeks now,
continuously monitoring and tweaking, and i have to reiterate what i
said in my first post to this group: i cannot quite believe how deeply
impressive this work really is.

i'm now at an incredible *0.8* seconds to main pageview and a 1.2 to
1.3 seconds to "all done" after the first page is loaded. the main
page even when pressing Ctrl-F5 and a completely cleared-down cache is
still around the 1 second mark on my laptop. bear in mind i'm in
taiwan and the server's in the UK, with a 300ms ping round-trip and
the times are even more impressive.

all the things that freaked me out such as "above the fold CSS" -
which i just... absolutely cannot be bothered to do *even if i knew
how to do it* - all taken care of automatically. now repeat that for
every single days-to-weeks-of-learning-and-implementing piece of
"trickery" that you have to do to get a responsive site... and it's
all done for you.

the one thing that has always pissed me off about doing web site
"optimisations" is that you have to maintain expertise in your head of
twenty, thirty, fifty different "tricks", but worse than that you make
one single core change to the design and you could end up having to
modify dozens of pages.

premature optimisation being the root of all developer evil, to have
an automatic highly-tunable tool which can do all the optimisation for
you.

an example: i can't tell you how many times i've resized and changed
the image quality of dozens of photos over the 10 years i've been
managing this site. i _was_ going to write an image cache, with
resizing so that pictures on the site so that they would be faster.
worpdpress... *shudder*.... i've seen *FIVE* copies be made (and
stored) for the exact same image as a "just in case" - so a site which
should take up 10 mbytes takes up a HUNDRED instead...

... now i can simply let pagespeed handle it... and it can work out
the best quality for the individual user.

also, i just wanted to say thank you to otto and joshua, because your
kind and patient questions and insights have prompted me to use the
right tools for investigating where time is being spent on loading,
even where pagespeed isn't actually involved (webpagetest.org for
example). i noticed for example that facebook badges take a whopping
780ms to be generated and served up, which is insane... so i simply
deleted it. i would not have thought to look closely otherwise.

also i always meant to investigate and deploy nginx some time: apache2
with mpm_event segfaulting prompted me to ditch it and use nginx
instead, which i'm now happier with - still have to get used to it,
but again i would not have considered using it without your insight
and expertise.

so anyway if there's anyone else on the team who's not posting on this
list, please do pass on my thanks to them.

all best,

l.

Joshua Marantz

unread,
Aug 25, 2017, 8:05:02 AM8/25/17
to mod-pagespeed-discuss
Great to hear, Luke!  We appreciate the sentiment.

Normally we hear only issues on this list (although those are good too because it means people are using the stuff).

-Josh


--
You received this message because you are subscribed to the Google Groups "mod-pagespeed-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mod-pagespeed-discuss+unsub...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/mod-pagespeed-discuss/CAPweEDws%3DC1oLH3LMNA7LEptcdMQQu5Bshd05vEhGREG%3D1xV_w%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Luke Kenneth Casson Leighton

unread,
Aug 25, 2017, 8:59:57 AM8/25/17
to mod-pagesp...@googlegroups.com
On Fri, Aug 25, 2017 at 1:04 PM, 'Joshua Marantz' via
mod-pagespeed-discuss <mod-pagesp...@googlegroups.com> wrote:

> Great to hear, Luke! We appreciate the sentiment.
>
> Normally we hear only issues on this list (although those are good too
> because it means people are using the stuff).

:)
Reply all
Reply to author
Forward
0 new messages