56k Times Five: Myth Or Moneymaker?

maxentius writes "InternetNews.com has an article on not-broadband-but-still-faster telephone internet access premiering soon in more than one commercial ISP venue. Compression and other techniques will improve speed by up to five times, so they say. Hi-tech or hogwash?"

Read the Article

[Obiwan Kenobi]

So the ISP will be using compression when a user requests a page or file. This won't help in the speed of downloading already compressed files, only web browsing.

Email speed will stay the same.

Downloading compressed files will stay the same.

Browsing will be somewhat faster, but 7x is a stretch.

More than anything, I bet most of those $28.95/mo customers will be paying for the privilege of ~5min support response calls.

Definitely file this one into the "Hype" category of Hogwash.

Won't work...

[ZorinLynx]

A normal telephone call, at least in the united states, is carried on a 64Kbps channel. IE, the sound is sampled, converted into a 64Kbps bit stream that is sent through the network to the other end, where it is converted back into an analog stream.

This makes it impossible to cram more than 64Kbps into a phone call. Sure, you can compress the data, but once data is already compressed (as images, movies, and other things people usually want fat bandwidth for), it can't be compressed anymore.

Unless they dramatically change the analog phone network, which won't happen, this is a pipe dream. Sorry guys.

possibly...

[ferrocene]

My ISP just implemented such a thing, and as an ex-employee I got to beta test. All the beta testers signed up for the new service as soon as the testing period was over, which is $5 more a month than the regular dialup. So it looks like they're doing something right.

Speed vs. Time

[ChrisKnight]

Just because web pages load five times faster, do not assume your connection speed is five times faster. The basis of the Plus service is a web optimization proxy server that sits between you and web servers. It automatically reduces the size on images, compresses the text, and does various tweaks to squeeze more into your 56k.

Your MP3s and bad porn will still come across just as slow on your gnutella client. Sorry.

-Chris

5:1 Compression...I Think Not

[VCAGuy]

I hope that EarthLink qualifies what they mean by 5x faster. They're probably talking about "user experience" speeds. Because, if you think about it, when we do backups, we use 2:1 compression as the "ideal," and everyone that's ever loaded Travan or DLT or DDS drives knows that when it says 200GB, it means 200GB compressed at 2:1. Short of some sort of very high-powered (in terms of CPU cycles) compression, 5:1 is almost impossible to achieve--certainly with desktop hardware, and probably not at all.

Re:Myth

[TopShelf]

They're using Propel's web accelerator. From Propel's website [propel.com]:

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

Re:I figured it out -

[Shishak]

Actually,

It saves the costly 3-way TCP handshake on the slow modem connection by installing a local side proxy. The proxy makes a couple permanent TCP connections to a squid proxy on the other end. I know for a fact propel uses squid on the server side. If the content is cached you save 1.5 * ping time to server for every request to that server.

How does it work?

[stratjakt]

From www.propel.com:

Propel Accelerator is designed to provide maximum acceleration for the Web sites you visit regularly.

So, the more you surf, the faster your favorite pages will load!

Specifically, Propel Accelerator speeds up the delivery of Web pages three ways:

Compression. Propel Accelerator delivers text and graphics more efficiently, using a proprietary compression technology that significantly reduces the size of Web pages and page elements sent to your browser.

Caching. Propel Accelerator intelligently retains and re-uses Web pages and page elements that have previously been sent to your PC. That's why the longer Propel Accelerator is in use on your PC, the faster your Web pages will load.

Persistent Connections. Propel Accelerator uses proprietary techniques to carefully manage and optimize the communication between your modem and our network of servers through a persistent connection. This eliminates the time wasted re-establishing and closing TCP/IP connections.
Looking for more technical detail on how Propel Accelerator works? Please refer to our Technical Overview. It explains the various components and how they interact with one another.

Nothing magic. It compresses a whole page, images and all, on the ISP side, and sends it down a persistant pipe to your client, along with some more intelligent caching information than is default (ie, the /. icons would stay cached but the text wouldnt).

It would probably 'look' faster since the whole page is delivered in one package, and renders all at once, rather than having text and waiting for images to show up.

It only accelerates HTTP AFAIK, so it's useless for anyone but the mom and pop web browser. It's certainly no substitute for bandwidth. The joe users buy broadband for P2P and streaming video and VPNs, none of which this 'technology' helps.

It also sounds like it would require client side software. Support? "Windows 98/NT 4.0/2000/ME/XP (sorry, no Macintosh support yet." which goes without saying.

Which brings me to a question. I regularly route my web browsing through my squid proxy at home (through ssh). Since my home uplink is 15k, it throttles my browsing. Is there an open source clone of this, or something similar?

Re:time to compress

[Jennifer E. Elaan]

When a web browser connects to a page, it (can) send a line called "Accept-Encoding" that describes what compressions it can understand. For instance, Konqueror sends "Accept-Encoding: x-gzip, gzip, identity".

Presumably, identity is standard uncompressed text. The others indicate its willingness to accept gzipped files from the webserver.

Since HTML is text, you have a GUARANTEE of 1/8th space savings. Since HTML tends to use a lot of similar codes, the space savings are, in all likelihood, far greater. Since on dialup, the latency of compression is trivial in comparison to the limitations of bandwidth, this may help substantially.

Web-server compression makes sense to me.

Then again, there are PPP extensions for compression now too. These would have a similar benefit.

Combined with both an off-site connection proxy and an on-site data proxy (this is what their webpage suggests they base their technology on), you get the enhancement they claim, more or less (not for compressed files or raw data transfer though).

Re:Myth

[reidbold]

You could try uninstalling the flash player.

Re:It's available now.

[Xawen]

Now if only the ISP would turn on compression on their end everything would be sweet (*).

Compression "enhancements" like this won't do you any good on your downloaded software or most images. Your downloaded programs are already compressed. Something like this can't crunch it much further, if at all. Pics like .jpgs are also pre-compressed (part of the format).

Compression works by eliminating repetitive data in a way that can be reveresed. You can only do it once. That's why you don't get a smaller file if you try to zip a .zip. So basically, these things are rip-offs unless you sit around downloading huge uncompressed files all day long.

JPEG2000 is done

[benwaggoner]

The standard is long since complete - ISO standard since December 2000. QuickTime for MacOS X has a good implementation of it. And yes, it has both lossless and lossy modes. And yes, the core coding scheme is license and and royalty free.

http://www.jpeg.org/JPEG2000.html

I'm really looking forward to JPEG2000 for digital cameras, since instead of having to cache thumbnails, applications like iPhoto can just decode the wavelet subbands appropraite for the current resolution. Much faster than having to decode the whole JPEG and then cache a thumbnail. Browsing an iPhoto library with 2000+ files strikingly slow, and surprisingly fast considering the math that is going into it.

Still, PNG will probably be better for synthetic graphics like screen shots, where JPEG2000 will be better for natural images.

Re:Myth

[Anonymous Coward]

Open regedit, and find:
HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Internet Explorer\ActiveX Compatibility\
Now look for a subkey called:
{D27CDB6E-AE6D-11CF-96B8-444553540000}
If it doesn't exist, create it.

Now create/edit a value in that subkey called Compatibility Flags, type DWORD.
Set the data in the value to 400 (that's as hex; 1024 as decimal).

Bingo - no more Flash, ever, and no more prompting to install the plugin either.

Also works for any other ActiveX control - if you know the CLSIDs that various spyware uses then you can block those too.

Re:Myth

[DietHacker]

This is not FUD but it is BS. Properly set your security settings and this will not happen often. Properly choose a proxy server and this ought not happed period. It really isn't a MSFT versus Unix thing. Educated user versus the other kind.

5x dialup speeds is based on 28.8kbps speeds

[t-maxx cowboy]

In Winnipeg, Manitoba, our DSL and Cable internet providers have a lightspeed, or lite speed. a.k.a. 5x dialup speeds. That 5x dialup is based on a 28.8kpbs modem, not 57.6kpbs modem. So yeah 12-15KBps is about the top end on those accounts.

Re:Myth

[Anonymous Coward]

After you uninstall the Flash player, search your hard drive for the file called "hosts" (no extension). It's buried in some subdirectory of your Windows directory. Open it in notepad, and add the line:

127.0.0.2 www.macromedia.com

and save it. From then on, you will never be nagged, and www.macromedia.com will be completely inaccessible in IE. Goodbye, Flash.

Re:I don't see...

[saider]

The 56k limit comes from the fact that the phone company's A/D converter is 8 bits wide with an 8kHz sampling period. The phone company also has the option of using your least significant bit for in-band traffic management (also called bit-robbing). This leaves you with 7 bits at 8kHz = 56k.

The modem is not the problem. It's the phone company and their standards.

Besides, modems already do this in analog mode (33.6k). The modem shifts the phase an frequency of the carrier signal to cram 30k of data into a 4kHz bandwidth. There's only so far that you can go.

ADSL modems are a different beast entirely and do not use the POTS (plain old telephone service) circuitry. They have a splitter that sends POTS traffic to the POTS circuit and the ADSL signals to an ADSL modem. They use the same wire, but they are separate signals to the phone company.

Re:Myth

[valmont]

No, actually, please do take a thorough read [propel.com] at all the technologies and methodologies involved in his acceleration framework. Compression is a tiny portion of many, many other things working together to speed your web surfing. Basically they're able to "isolate" static and/or non-changing portions of documents and "cache" those portions locally on the hard drive. Their local proxy maintains persistent connections to remote proxy servers, which can save tremendous time typically spent in DNS lookups and TCP SYN/ACK transactions during initial stages of an HTTP transaction.

it's actually pretty bad-ass.

Re:Hmmm

[ScottKin]

Unfortunately, all of this "dark fiber" that is so frequently spoken of is *not* FTHoP (Fiber To Home or Premises), but over-ordered and unlit fiber between MANs (Metropolitan Area Networks). A backbone provider worth their salt will always over-order fiber strands due to fiber-breaks, additional fiber pathway protection and redundancy-on-top-of-redundancy. SONET BLSR (Bi-Directional Line-Switch Ring) takes x2 of the needed fiber strands because you must have an alternate fiber path to provide circuit/ring redundancy to cover fiber breaks/cuts, where as UPSR (Unidirectional Path-Switched Rings) only needs 1 strand per fiber ring because fiber breaks/cuts or failures will cause the equipment to switch-out and "wrap" the ring to try to keep some sort of integrity on the ring and to try to minimize the number of nodes switched-out of the ring.

FTHoP won't be a reality in most neighborhoods for some time to come because of the exorberantly-high prices - unless the city has been forward-thinking enough to include fiber networks pre-built into the city's infrastructure. "Dark Fiber" is a misonmner and does not include FTHoP facilities.

Sorry to burst the bubble - but demz da facts.

ScottKin