RSS is the Web 2.0 Pipe

作者：Dion Hinchcliffe 日期：2006-3-1

What did I miss? What other canonical examples are there?

Technorati: web2.0, rss

Hooman made this comment,

RSS is undoubtedly an important first step on the path towards an interconnected network of shared social services. However, I think that everyone needs to step back from and take a hard look at it from a software design perspective. RSS is not and, frankly, should not be the mechanism to "tie" a Web of data together. There is no way to discern the structure, or meaning of the data in RSS 2.0. Yes, there is more structure than a simple text page. However, how do you know what the content contained within a feed actually means? By the title? By the channel name? By the item name? How do you know what an item name means relative to other items? How do you put the data in context to be manipulated in a useful manner? Where are the objects? For those of us that are data junkies and subscribe to 100s of feeds, this is a serious problem. It is great that I do not have to hope through websites to find information. But I still am constrained by the same keyword search mechanisms that I find on the web. Bottom line, although RSS is a good first step, there is a lot more to be done. Keep hacking folks.

David Gibbons made this comment,

Hi Dion.

I agree on syndication but don't share your enthusiasm for RSS other than that.

RSS is an unscalable, latent and highly ineffecient technology to impliment for either "Notification" or "Glue" - frankly, it's bad advice to apply RSS to those two use cases.

The reason - the business processes requiring N or G are event-dependant - RSS is a "pull" protocol - events require push + pull to be properly orchestrated.

Applying RSS to N & G will result in one of the following undesiderable outcomes ... 1) latency (this is why you have to ping technorati) or .... 2) VERY high and technically redundant processing overhead (this is why they don't ping you more often)

These are hard lessons learnt in enterprise internet trenches --> pub-sub doesn't hack it for Enterprise business services - other than syndication. RSS IS "really simple" and should not be applied to all integration problems.

The closest the web has to a standard today that I would advise using for N or G is ebXML - but few RSS hackers will be able to groc. it (myself included).

Dion Hinchcliffe made this comment,

David,

I would have agreed with you in the past but then I appreciated the numbers and the scale. In my post, I clearly stated that RSS is good enough 9 times out of 10. Given where it will be used and who will be using it (tens or hundreds of millions of uses), I bet I might have gotten the proportion wrong. I think RSS is probably appropriate 95 out of a 100 times or more in most given Web service situations. When it's not, I believe you should use the proper tool for the job, whether that's POX/HTTP, REST, SOAP, WS-*, JSON, or whatever.

And yes, high latency is an unfortunate side effect of pull only protocols. For many applications, that's actually acceptable, and so far pinging has been a workable reinforcing adjunct. When it's not, I agree that you have to move to something else. See my next post about "tolerance continuums", which I think addresses some of your concerns in more detail.

Best,

Dion

本新闻共2页,当前在第2页 1 2

文章来源：web2.wsj2.com

责任编辑：老冬瓜