Note: This is an internal discussion about subscriptions, rather than a reference to Blogrolling.
[BillSeitz] A blog roll is a list of a blogger's "favorite blogs" (and other sites?).
Should this become the same as someone's list of EntryChannels subscriptions?
If we think of people subscribing to hundreds/thousands of channels, what does that mean? Maybe have a "favorites" list which becomes the blogroll, as a subset of the subscriptions? Listing hundreds/thousands of sites on a front page seems silly - maybe keep those in a separate file?
Some services already track similarities across blogroll lists, so reducing the need for HTML scraping seems like a good thing.
Should each site publish this as a separate file, in addition to being part of the front page?
If available as a separate file, should it be in OPML, some other new standard format, or does it matter?
Within the front page, does it need a special format? Should it simply be contained within "a name=blogroll" (or rather, "div class=blogroll") tags?
[AlexanderPayne + PhilLeif RefactorOk] We'd like to see an element in the syndication file that links to an OPML file (the present de-facto standard for "exported" blogrolls). This would be totally optional, of course, as not everyone has a blogroll or other list of subscriptions, but it would be a Good Thing. Shoving a big list of links in the syndicated feed would probably turn most people off of this idea.
[ZhangYining RefactorOk] I think it's not part of syndication, maybe a standard API method signature is all we need.
[DanDickinson RefactorOk] So then: Element named blogroll, optional, type URI, should link to an OPML file? Sounds reasonable to me. Would be great for auto-discovery.
[DanDickinson] On second though, after reading the comment on OPML, maybe we should allow for a type specification, thus allowing people to use whatever blogroll format they'd like (such as OML)?
[BillSeitz] I don't see why OPML or OML is needed for something as simple as a blogroll. It's not a tree, it's an ordered list with 2 properties: label and URL. Why not keep it simple? It doesn't even have to XML. (I can imagine reasons to have richer info in a blogroll, like categories or ratings, but I don't see that as an area bloggers want to put effort into.)
[AdamRice] Lots of people already have subscription lists (which, as FrancoisPlanque points out, are very similar) in OPML, and I have seen blogrolls with 2-level hierarchies. The overhead for supporting OPML doesn't seem like it'd be very great. This does open up a can of worms, though, since the next step is including OPML snippets of last 5 tracks listened to, last 5 books read, etc. In a sense, this seems contrary to the general spirit of !Echo, which is to store and transact author content; in another, it is in keeping with !Echo's spirit, which is to comprehensively cover what blogs (and other future "microcontent" formats) are doing. Providing a standard, generic format for nonauthored TasteData (I just made that up) could open the door to interesting taste aggregation tools and the like. In that respect, I think the pros outweigh the cons.
[DannyAyers RefactorOk] I strongly believe that for this and similar purposes (such as Foaf) we need to have a solid SyntaxExtensionMechanism in the core. It should be possible to link in an existing OPML but for reasons given on that page I think it would be a really bad idea to tie blogrolls to that format.
[FrancoisPlanque RefactorOk DeleteOk] I think that a blogroll, as well as a subscribed feeds list, should actually all be considered the same: these are all sub-channels of sub-items. Those all have titles, links, description, categories... The only difference is the "class" of the chanel (posts, blogroll, subscriptions...) and their sort order (posts are typically sorted by date, blogrolls by name or by category and name...)
I think it would be useful to have echo support a tree structure of channels. Each channel could either be included inside the current file or referenced externaly by an URL. Regarding the blogroll, it could be included in the posts echo feed if it's small or in an external file if its long. Then each of the blogs in the blogroll that has an echo feed of its own would also be linked at...
Just my 2 cents... but I could go into much more details if some of you like this concept as much as I do
[JonathanSmith RefactorOk] Sorry, Bill, just got around to thinking about your questions. Looks like a nice web services opportunity, perhaps an adjunct to Feedster. A database of weblogs exists and someone wants to add some weblogs and build an OPML, so that person goes to the database and examines a list of categories.
Say, for instance, you are interested in gardening. You click the gardening category and several avid gardeners have listed their OPMLs with descriptions. You read the descriptions and select an OPML to import. Certainly, this is a recycled idea, rather than something new. It would be similiar to the way Amazon allows its "members" to build booklists or the way that AmphetaDesk helps a new user build an OPML.
[AsbjornUlsberg] I think that we either should just leave BlogRoll out of Echo and label as an extension (somehow), or we should define an entirely new format. The OPML format is just horrific, and I at least don't want to see it in Echo in any way. I think XBEL is a much better format to "standardize" upon or at least just to get ideas from. OPML should imho just be forgotten as soon as possible, except it's functionality.