Date: Wed, 24 Jun 92 12:02:25 EDT From: davis@willow.tc.cornell.edu (Jim Davis) Message-Id: <9206241602.AA13181@willow.tc.cornell.edu> To: www-talk@nxoc01.cern.ch Subject: Raisch's Attention link Cc: <paragraph> In a previous message, R. Raisch proposed a number of interesting link types. One of them was the attention link, which I want to discuss here. I am uncertain about the need for this link, the technical ability to provide it, and the definition. <nl> <paragraph> Let me discuss the function first. The function seems to be providing a means of notifying an "owner" of a document when certain conditions obtain. The conditions Raisch mentions are:<nl> <indent> 1) someone has read the document<nl> 2) someone has modified the document<nl> </indent> <nl><paragraph> There is some doubt in my mind about read notifications. I am not sure I want other people to know what I read. I can see only two reasons for this feature:<nl> <indent> 1) as a form of "proof of delivery". Some email systems provide this, but I don't like it.<nl> 2) as a means of collective revenue (pay-per-read)<nl> </indent> <nl><paragraph> As for charging, fairness requires that the link not be activated until I have seen a warning (otherwise I might get charged a zillion dollars to read the document - just like 900 phone numbers in the USA). So this will add complexity to the client. <nl><paragraph> Also, attention links are not sufficient for a charging. They support a model where I am charged once per read, no matter how much of the document I read. But it seems likely that there might be need for other charging models. It is also unclear that they are technically sufficient. For economic purposes, you would want the message to be sent by the server, not the client. But this would cease working if the document were copied. (But maybe this is not a fair objection, since I know of no scheme that can preserve property rights given the possibility of perfect digital copying.) <nl><paragraph> Continuing on the more general question of read notification, regardless of purpose, it is possible that one might desire notification on a finer grain that the entire document. But this, I think, requires the cooperation of the client. Indeed, the client can tell the server that a given piece of text has been displayed, but not whether the user actually read it (unless we go further, and implement it with an executable function which requires the user to click on a button) <nl><paragraph> As for the second form (modification notification), it seems to me that there is a need to inform not just the owner, but also other people. There are two reasons for this: <nl><paragraph> First, as owner of a document, I am not likely to allow other people to modify my document at all. On the other hand, I might be interested in notifications when someone adds or deletes a link <bold>to</bold> my document. But attention links don't address this problem. <nl><paragraph> Second, as a reader of a document I don't own, I might want to be notified when the owner modifies it, since I might wish to re-read it (or at least the changed sections.) Let's call these "monitor" links. Monitor links might be a useful means of reducing effort required for some kinds of network retrievals - those where I am interested in new developments in certain areas. Now, instead of polling documents to see whether they've changed, I can just leave an "attention link" and get a notification. <nl><paragraph> On the other hand, this has some problems. One of them is the question of who pays the cost of sending all these notifications. Can you imagine the load on your workstation as it sends out 10,000 monitor notifications? Perhaps this can be answered by bringing in more economics - that is, to attach a monitor link I need to set up an account such that I can be charged for the delivery. Or maybe the notifications are sent by the document server, so as an author I am not affected. <nl><paragraph> A second problem (or at least issue) is that monitor links require finer grain of size and time. Some users will want to monitor only select portions of a document. Likewise, we may not want notifications sent when <italic>any</italic> editing is made, but rather only when the author completes a session. That is, if I edit the document for a day, saving changes six times, you don't want six notifications. This might require some notion of "transactions" such as used in data bases. <nl><paragraph> Finally though, I don't think it is correct to call these things links. They are not related to logical structure, nor are they explicitly activated by the reader (or writer), indeed that person not even be aware that it was activated. Consider the more general question - if attention links are a subcase of execution links, and attention links can be activated without knowledge of the user, should all execution links be capable of such activation? Do you want to read documents which can cause arbitrary computations to occur without your choice, or even knowledge?