Don't bother saving white-space

[Stephen Poley]

People in these groups, and on web-pages, not infrequently suggest that
it is worthwhile cutting down on white-space and comments in HTML and
CSS in order to reduce loading times. I and others have more than once
doubted this, given the data-compression in the HTTP protocol. Having
seen it suggested again a couple of times in the last few days, I
decided it was time for a test on the effect of white-space.

I took one of my pages:
http://www.xs4all.nl/~sbpoley/webmatters/checklist.html
which is 20 Kb.

I then bloated it with whitespace to 162 Kb (nothing special about that
number - it's just what it happened to end up as):
http://www.xs4all.nl/~sbpoley/misc/checklist.html

I tested them over my 46 Kbps modem connection (yes: 46, not 56; don't
ask - I don't know either) using Opera 7.

The first page loads in 3 seconds, the second in 8 seconds - both
figures seem to be repeatable. This suggests that if you took a file
with a fairly generous 5Kb of white-space, and stripped out all of it,
loading would be speeded up by a princely one-sixth of a second. (For
comparison, the largest HTML file on my site, of 79Kb, turned out to
have just under 3Kb of compressible white-space.)

Somehow it just doesn't seem worth it ...

--
Stephen Poley

http://www.xs4all.nl/~sbpoley/webmatters/

[Neal]

On Thu, 19 Aug 2004 20:49:58 +0200, Stephen Poley
<sb************ ******@xs4all.n l> wrote:

People in these groups, and on web-pages, not infrequently suggest that
it is worthwhile cutting down on white-space and comments in HTML and
CSS in order to reduce loading times. I and others have more than once
doubted this, given the data-compression in the HTTP protocol. Having
seen it suggested again a couple of times in the last few days, I
decided it was time for a test on the effect of white-space.

I took one of my pages:
http://www.xs4all.nl/~sbpoley/webmatters/checklist.html
which is 20 Kb.

I then bloated it with whitespace to 162 Kb (nothing special about that
number - it's just what it happened to end up as):
http://www.xs4all.nl/~sbpoley/misc/checklist.html

I tested them over my 46 Kbps modem connection (yes: 46, not 56; don't
ask - I don't know either) using Opera 7.

The first page loads in 3 seconds, the second in 8 seconds - both
figures seem to be repeatable.

For me, original=3secs, bloated=21 secs.

The extra white space amounts to 142 Kb. Took me 18 extra seconds to load.
About 8Kb per sec, or close to the same for the original file. This would
seem to indicate there is value in reducing white space.

[Stephen Poley]

On Thu, 19 Aug 2004 15:00:09 -0400, Neal <ne*****@yahoo. com> wrote:

On Thu, 19 Aug 2004 20:49:58 +0200, Stephen Poley
<sb*********** *******@xs4all. nl> wrote:

The first page loads in 3 seconds, the second in 8 seconds - both
figures seem to be repeatable.

For me, original=3secs, bloated=21 secs.

The extra white space amounts to 142 Kb. Took me 18 extra seconds to load.
About 8Kb per sec, or close to the same for the original file. This would
seem to indicate there is value in reducing white space.

That's interesting. I wonder what causes the difference? What browser
are you using? Are you using a 56Kb modem? - if so, it indicates that it
must be doing some compression, or the 'bloated' transmission would take
around 30 seconds.

--
Stephen Poley

http://www.xs4all.nl/~sbpoley/webmatters/

[Els]

Stephen Poley wrote:

On Thu, 19 Aug 2004 15:00:09 -0400, Neal
<ne*****@yahoo. com> wrote:

On Thu, 19 Aug 2004 20:49:58 +0200, Stephen Poley
<sb********** ********@xs4all .nl> wrote:

The first page loads in 3 seconds, the second in 8
seconds - both figures seem to be repeatable.

For me, original=3secs, bloated=21 secs.

The extra white space amounts to 142 Kb. Took me 18 extra
seconds to load. About 8Kb per sec, or close to the same
for the original file. This would seem to indicate there is
value in reducing white space.

That's interesting. I wonder what causes the difference?
What browser are you using? Are you using a 56Kb modem? -
if so, it indicates that it must be doing some compression,
or the 'bloated' transmission would take around 30 seconds.

I'm on broadband, the first is done within a second, the second
one takes 2/3 secs.

I suppose that could add up if you're working with 'flip-
through' pages, even for broadband.

--
Els http://locusmeus.com/
Sonhos vem. Sonhos vão. O resto é imperfeito.
- Renato Russo -
Now playing: Magnum - No Way Out

[Spartanicus]

Stephen Poley <sb************ ******@xs4all.n l> wrote:

People in these groups, and on web-pages, not infrequently suggest that
it is worthwhile cutting down on white-space and comments in HTML and
CSS
I'd suggest that whitespace and comments are 2 different things.
in order to reduce loading times. I and others have more than once
doubted this, given the data-compression in the HTTP protocol. Having
seen it suggested again a couple of times in the last few days, I
decided it was time for a test on the effect of white-space.

I took one of my pages:
http://www.xs4all.nl/~sbpoley/webmatters/checklist.html
which is 20 Kb.

I then bloated it with whitespace to 162 Kb (nothing special about that
number - it's just what it happened to end up as):
http://www.xs4all.nl/~sbpoley/misc/checklist.html
What you've done there is not typical coding whitespace, usually this
refers to blank lines (cr/lf pair on a line of it's own) and tabs, not
padding a document with spaces.
I tested them over my 46 Kbps modem connection (yes: 46, not 56; don't
ask - I don't know either)
56K is the theoretical maximum, in reality this is never achieved, sub
50K is pretty normal.
The first page loads in 3 seconds, the second in 8 seconds - both
figures seem to be repeatable.

I don't know anything about http protocol data compression, but note
that modem connections typically include the V42b protocol. V42b
provides real time compression up to 4x. Ascii data typically compresses
very well.

--
Spartanicus

[Neal]

On Thu, 19 Aug 2004 21:39:41 +0200, Stephen Poley
<sb************ ******@xs4all.n l> wrote:

On Thu, 19 Aug 2004 15:00:09 -0400, Neal <ne*****@yahoo. com> wrote:

On Thu, 19 Aug 2004 20:49:58 +0200, Stephen Poley
<sb************ ******@xs4all.n l> wrote:

The first page loads in 3 seconds, the second in 8 seconds - both
figures seem to be repeatable.

For me, original=3secs, bloated=21 secs.

The extra white space amounts to 142 Kb. Took me 18 extra seconds to
load.
About 8Kb per sec, or close to the same for the original file. This
would
seem to indicate there is value in reducing white space.

That's interesting. I wonder what causes the difference? What browser
are you using? Are you using a 56Kb modem? - if so, it indicates that it
must be doing some compression, or the 'bloated' transmission would take
around 30 seconds.

Opera 7.23 on dialup 56k, 45333 bps.

[Alan J. Flavell]

On Thu, 19 Aug 2004, Stephen Poley wrote:

I took one of my pages:
http://www.xs4all.nl/~sbpoley/webmatters/checklist.html
which is 20 Kb.

I then bloated it with whitespace to 162 Kb (nothing special about that
number - it's just what it happened to end up as):
http://www.xs4all.nl/~sbpoley/misc/checklist.html

It may be of interest that when gzipped, your file sizes came out as
7711 and 8805 respectively. So as we've been saying all along (well,
since NCSA X Mosaic had been supporting gzipped HTML about a decade
back, courtesy of jwz), there's far more to be gained by serving out
HTML gzipped than by fussing about a bit of white space.

thanks for the heads-up, though - no offence meant! ;-)

[Andrew Graham]

Stephen Poley wrote:

last few days, I decided it was time for a test on the effect of
white-space.

I took one of my pages:
http://www.xs4all.nl/~sbpoley/webmatters/checklist.html
which is 20 Kb.

I then bloated it with whitespace to 162 Kb (nothing special about
that number - it's just what it happened to end up as):
http://www.xs4all.nl/~sbpoley/misc/checklist.html

I tested them over my 46 Kbps modem connection (yes: 46, not 56; don't
ask - I don't know either) using Opera 7.

The first page loads in 3 seconds, the second in 8 seconds - both
figures seem to be repeatable. This suggests that if you took a file
Here are two more data points for you, both over 24kbps modem from USA
west coast:

1) Firefox:
9 seconds vs. 13 seconds

2) wget:
D:\test>timethi s wget -q
http://www.xs4all.nl/~sbpoley/webmatters/checklist.html | grep "Elapsed"
TimeThis : Elapsed Time : 00:00:05.107

D:\test>timethi s wget -q
http://www.xs4all.nl/~sbpoley/misc/checklist.html | grep "Elapsed"
TimeThis : Elapsed Time : 00:00:15.322

Living in an area without high-speed connections, I appreciate more than
most the value of fast websites, but I still indent and wrap my html.
Using your page as an example, it is 19135 bytes without any newlines or
duplicate spaces, and 20803 bytes indented and wrapped at column 72.
Tiny difference.
Somehow it just doesn't seem worth it ...

Yup. Optimize those jpegs instead.

[Arne]

Stephen Poley wrote:

People in these groups, and on web-pages, not infrequently suggest that
it is worthwhile cutting down on white-space and comments in HTML and
CSS in order to reduce loading times. I and others have more than once
doubted this, given the data-compression in the HTTP protocol. Having
seen it suggested again a couple of times in the last few days, I
decided it was time for a test on the effect of white-space.

I took one of my pages:
http://www.xs4all.nl/~sbpoley/webmatters/checklist.html
which is 20 Kb.

I then bloated it with whitespace to 162 Kb (nothing special about that
number - it's just what it happened to end up as):
http://www.xs4all.nl/~sbpoley/misc/checklist.html

I tested them over my 46 Kbps modem connection (yes: 46, not 56; don't
ask - I don't know either) using Opera 7.

The first page loads in 3 seconds, the second in 8 seconds - both
figures seem to be repeatable. This suggests that if you took a file
with a fairly generous 5Kb of white-space, and stripped out all of it,
loading would be speeded up by a princely one-sixth of a second. (For
comparison, the largest HTML file on my site, of 79Kb, turned out to
have just under 3Kb of compressible white-space.)

Somehow it just doesn't seem worth it ...

My opinion is that the total size (i.e. images included) is to be
considered.

According to http://www.websiteoptimization.com/ a analysis on your page
/webmatters/checklist.html gives this result

Total Size: 43638 bytes
HTML: 19685
Images: 18539
CSS: 5414
Total Images: 4
Download Times*
56K 8.90 seconds
Same analysis on the page misc/checklist.html

Total Size: 189818 bytes
HTML: 165865
Images: 18539
CSS: 5414
Total Images: 4
Download Times*
56K 38.03 seconds

So, if (at it looks) you don't change anything else than adding
whitespace to the code, the download time change with aprox 29 sec for a
56K dial up connection!

--
Arne

[Spartanicus]

Spartanicus <[e-mail] > wrote:

http://www.xs4all.nl/~sbpoley/misc/checklist.html

What you've done there is not typical coding whitespace, usually this
refers to blank lines (cr/lf pair on a line of it's own) and tabs, not
padding a document with spaces.

A more realistic example of documents with and without coding
whitespace:
http://www.pan-europe.utvinternet.ie...hite_space.htm
(50,066 bytes)
http://www.pan-europe.utvinternet.ie...hite_space.htm
(48,658 bytes)

In addition to the tab's and cr/lf's that I normally use I've added a
blank line between every line of code in the first document.

--
Spartanicus