Curtis, I think that this discussion should continue for its instructive merit, but I think that it has ceased being a debate about the rate at which the Public Internet continues to grow. At least this should be the case if you think it appropriate to add other froms of networking to the bundle.
Once we depart from the domain names of IANA, and the routing and directory rules that govern them, and begin to introduce arguments that include extranets and intranets, we are simply discussing any and all forms of networks that use, somewhere in their make up, TCP/IP, despite the fact that they are actually tied to the 'net at some point through gateways. These, I maintain, do not the Public Internet make. At least not under the semantics we've [I've?] grown accustomed to.
Those network forms, to be sure, would soon be near all-encompassing, since most networks going in today are IP based on some level, and many of yesterday's legacy nets, including SNA, are now being transformed through IP on the Main Frame modification techniques, and SNA over IP over private WANs.
Once you include enterprise legacy networks which are being transformed to TCP/IP support, especially those which use IP gateways to supply chain organizations and customers, then the proposition that these intranets and extranets should be counted in what actually goes into the growth rate of the Public Internet is clearly inconsistent with commonly held beliefs as to what the Public Internet actually is.
True, these enterprise nets [of many different flavors] are at the same time opening up to inside and outside user pools through the use of IP (and other forms of) gateways, very often from the Public Internet. This, plus the use of DMZs and proprietary security measures in most cases, only serves to differentiate them, IMO, rather than to include them as a part of the Public component.
To drive the point, let me ask: If a Fortune 100 company with a flow of 400 Terabits per day of image archiving activity, opens up to a select group of users from the 'net who want access to their cancelled checks, using gateways and security measures, say a dozen or so highly valued customers, does this then enittile the entire lot of 400 Terabits per day to get lumped into the 100-day run rate? Of course it doesn't. Or does it, in your estimation?
Stated another way, while they often use the same IP protocols as the Public Internet, especially at the distribution levels [at the LAN level to end users], and are most often tied to the Public Internet through gateways, that's where both their similarities end, and their distinctions begin.
Extranets are not tied to individual, discrete technologies. Not even TCP/IP, as I think you are implying they are due to their heavy dependence in most cases to IP. Rather, they are platforms made up of many different technologies and may consist of many different transport protocols.
Extranets might employ dedicated SONET rings and ATM when deemed necessary, even without IP riding in the caboose, at Layers One and Two for bulk data transfers for example, in segmented fashion.
They will use Frame Relay when such is in-place and leveragable. They use their own Class A and B address ranges and subnets [which exempts them from the scaling problems associated with normal Public Internet routing parameters, i.e., they scale to their own unique set of internal parameters.
These mutated network forms are, in fact, universes of their own order which can be compared to, and sometimes tie into, but are not the same as, the Public Internet.
Different protocols are mixed, different sets of DNS rules apply, different policy agendas, different security measures, different directory schemes are used, a different a lot of things prevail.
In time, most enterprises will extend their internal databases and business processes via TCP/IP networks to their industry partners, indeed many, if not most, of these will even use partitioned bandwidth made available by ISPs. But that does not make them a part of the Public Internet by any stretch of the imagination.
We are, of course, arguing a soon-to-be moot point, IMO, as the Public Internet, as we've known it historically, can only last so long without changing its identifying traits and characteristics. But if changing those characteristics means shedding some of the qualities and attributes, while adding many others, thus changing its persona from what it has been thus far, then I think that we'd better find a new word to identify it [or them]. Which in a way is precisely what we've done, I think, by adding the categories of intranets, extranets, etc.
At the same time, the bit counters who find it beneficial to their cause to announce sensational numbers had better find a better way to express what it is, exactly, that they actually mean. And while they are right about one thing at the very least, they have instead lumped all of the above into a single category, and with the exception of not meeting academic clarity, they are probably right with their extrapolations and predictions to within tolerable margins of error.
But I maintain the distinctions I've made throughout this discussion pertaining to the original argument. Which all serves up the point rather nicely that we can say the same thing different ways.
---
I found one of the points you made in closing very interesting:
>>Of course, these growth rates cannot continue forever--there are only so many people on the planet and each can only use so much bandwidth. Currently, the number of people that use it is growing, the uses for it are expanding, the number of network intelligent boxes is increasing. As a result, the traffic grows--doubles every 100 days or so. I believe the financial community knows this.<<
I agree that one of the dominant factors here thus far has been the number of people available to populate the 'net, no doubt. But I also think that once capacity issues are improved dramatically, which they inevitably will [but only in relative terms], the next set of drivers will be the applications that those users subscribe to and use, which will account for an even greater level of next-gen demands, testing the ability of the 'net to deliver on its promise.
It will not, in other words, see any abeyance, solely from the diminishing supply of additional users.
Any comments on this? Will saturation of end points mark an end to the demand burden, or even have any material impact on it? Or will incremental applications, the now-famous "application creep," keep the purveyors of bandwidth busy through subsequent generations? Or, will we see a form of "topping out," as my iron worker friends like to say, at some point.
Best Regards, Frank Coluccio |
