Slashdot Mirror

This is nothing new; it sounds like the same thing the "download accelerators" have been doing for years. My ISP has been offering Propel for almost 5 years. The only difference is that now a browser vendor gets to collect stats about your web use.

You make a very valid point, but why limit yourself to only using the modem? If we have an ISP at one end, and an end user with a PC (using a modem) at the other, could you not simply grab the data from the web at the ISP side, compress the entire file (be it html, jpeg, gif, zip) and then send the compressed version over the phone lines to the end user where the PC does the decompression (not the modem)??? Sure some things compress better than others, and streaming media such as audio and video could not use this technique, but the speed at which modem users could browse the web could be vastly improved.

This has been done. By many people. See here for one example.

Instead of reading the article, read Propel's technical specs, they are far more informative and insightful. You will that there is much, much more involved than simple compression. Also this technology is not meant to improve the data transfer speed of all protocols (POP, SMTP etc.), it is designed to specifically improve HTTP transactions during typical web surfing.

enjoy.

it's actually pretty bad-ass.

• persistent connections. that'll save you time wasted on redundant name lookups, and initial server connections.
• caching at various layers, including individual elements of pages, using dynamic reconstruction of pages while isolating updated content.

Simply brilliant.

Random ASCII data is not a text file, it's a binary file. A text file typically only contains the characters you can type with a keyboard, plus a few special characters like carriage returns and tabs. Also, most text files contain words primarily in one language which causes some characters to appear much more frequently than others. This allows those frequently used characters to be represented by only a couple of bits rather than an entire byte.

Most text files compress extremely well, I frequently see text files that are compressed to roughly 20% of their original size.

I'm skeptical of their ability to significantly compress graphics and other multimedia components of pages because they're usually already compressed.

Now, as for the technology as a whole, if you go to propel's general Technical Overview, You'll also see that it's not just a compression technology, it's also proxying and caching technology. They have a local http proxy with a persistent connection to their remote proxys. This should also give a small performance boost.

To me it sounds like it could significantly increase web browsing speeds. I just don't think it's worth paying the extra cash if you can get broadband in your area for a few bucks more per month.

One thing that I think they should implement if they haven't already is predictive caching. They should try to guess where your next click is going to be and start downloading that content to your proxy in advance of you hitting it. This can be especially effective in an environment with a large userbase where they can predict your next page based on other users' behavior.

We're doomed.

What will be accelerated
All text - HTML, markup, and javascript
Most graphics & photos - including jpeg and gif images and most Flash images and animation
Most banner ads
All browser-based emails
All emails that contain images - even when read in a dedicated email program

What will *not* be accelerated
Streaming media, and audio and video files
Secure pages, such as those used for online banking and credit card forms
MP3 files and executable programs

The article links to Propel Software. On their front page they claim their product can give a 5x speed improvement to a normal dial-up account.

This is not Earthlink providing new connection technology; it is simply a proxying/compression trick, hardly worth $7US a month.

earthlink uses propel accelerator, some program (clientside and isp-side?) that optimizes your connection (caches webpages, finetunes modems' settings).

nothing very exciting, this is still our good old 56k technology in action.

faster 56k's are still a myth.

but this offer, if Joe Modem turns out to be dumb enough to believe it's real, is a moneymaker.

If you rmotfa you note that they partner with propel web accelerator. If I remember correctly there was some talk on /. earlier about them.

You find out that Earthlink isn't actually changing the dialup speed at the modem level .. they are just reselling Propel software's Accelerator product. Earthlink is charging a $7/month premium over their standard dialup, so Earthlink subscribers get a full $0.95 / month savings over simply buying Propel's offering.

You find out that Earthlink isn't actually changing the dialup speed at the modem level .. they are just reselling Propel software's Accelerator product. Earthlink is charging a $7/month premium over their standard dialup, so Earthlink subscribers get a full $0.95 / month savings over simply buying Propel's offering.

Clientside caching surely is most of the speed.
Serverside caching could be used.
TCP/IP non-comformaty is the third option.

Assuming this is true, (ignoring the 1500 lines a day), what else could he be doing?

Judging by harddisk prices, client side cacheing algorythms would make sense. Cacheing many portal and search engine homepages is a powerful start. Combined with a central server that then reviews these popular pages for changes, and publishes a simple summary for the browser client to collect and compare with older summaries, then a browser can collect only updated portal pages for the cache, all optimizes portal renders.

Then less common homepages, such as the high school I attended, can be gleened from users typed-in webaddress history, and automatically cached as chron-job.

Creating cached copies of commonly used graphics on portal website can save a ton of bandwidth. Again a server based bot could rate the linkcount of graphics on portal sites, and if the graphic has changed, and then post this list for browsers to collect for caching. Searching HTML for imagefiles, that are already stored in the cache, and modify the page on the fly to call only the cached image would save bandwidth. e.g. caching all of slashdot's article catagory icons.

Then the tricky part, "which linked pages to cache while the user reads a page?", so that when a link is clicked, the pages renders fast. I would download the html from all of them, and while the reader reads, check for already cached images, and then start downloading image files.

-Mac Refugee, Paper MCSE, Linux Wanna be!
and first poster of the word "knoppix"