October 19, 2005
@ 04:56 PM
Comments [12]
RSS Bandit [Nightcrawler Edition] Update: Amazon OpenSearch, the Newsgator API and a New Logo

It's been over a month since we shipped the alpha version of the Nightcrawler release of RSS Bandit. Since then we've fixed a number of annoying bugs and polished a number of our features. An example of the kind of polish we've added since the alpha is shown in the screenshot below.

There are three main classes of subscriptions we now support in RSS Bandit; feeds (Atom or RSS), newsgroups (NNTP) and search results. We made search results a first class subscription type because I suspect that subscribing to search results especially on various blog search engines is only going to increase in popularity. The process for adding a new search engine is still too "techie-focused" for my liking. I'd love it if our users would just be able to add the URL of their search engine of choice and then we check if it supports Amazon OpenSearch, if so then we add it as one of the choices for the Search Results subscription wizard. The current process for adding a search engine whose results can be subscribed to involves users adding a URL showing the query string format of the engine (e.g. http://search.msn.com/results.aspx?q={0}&format=rss where {0} is a place holder that shows where the query string should be inserted).  

I should investigate how many search engines provide an OpenSearch description documents file. If enough of them do, it may be worth the while for our users if we went ahead and supported it. That way they can just add 'http://search.msn.com' to their favorite search engine list and we autodiscover the rest.

The Newsgator API has been a source of mild frustration for me since I added support for it. The existing synchronization features in RSS Bandit involve uploading/downloading a single file containing the state of the application. The Newsgator API assumes that any application using it for synchronization is also using it as a source for RSS feeds. From my perspective this seems to be a very big assumption to make but is understandable when one considers that the original purpose of the API was for their in-house applications. This assumption manifests itself by requiring that to synchronize the state of a feed I'm subscribed to, I need to fetch its feed from Newsgator online. This means that if I'm subscribed to 100 feeds in RSS Bandit, then I might need to download up to 100 feeds from Newsgator Online as part of the synchronization process each time I sync. This makes the synchronization process a lot slower than I expected. I'm now wondering whether we should rethink the user flow for our synchronization step since currently we lock the UI while syncing to prevent users making changes while we are syncing. With synchronization to Newsgator this could take several minutes as opposed to a minute or less with our other synchronization methods. I did make some performance improvements since we shipped the alpha but it still does take a while longer than I like. :(

The winner of the RSS Bandit New Logo design contest has been announced. Congrats to Eric Winchester.

Old Logo:

New Logo:

I'd like to thank all the folks who took the time to submit entries and those who voted for their favorite logos. We greatly appreciate your support.

It is quite likely that Torsten and I will ship a 'refresh' of the alpha installer this weekend. The reason it isn't a beta is that we are not yet feature complete. The code for downloading enclosure/podcasts still isn't all there and I haven't started on my idea for 'watching' posts for new comments. With any luck we should have all this done in the next few weeks.

After the beta, we'll focus primarily on performance issues. We've already fixed a number of issues that were causing lots of CPU usage but our memory consumption still higher than I prefer. I expect that the final version of Nightcrawler will ship during the holiday season.


 

Categories: RSS Bandit

« The Myth of the Office XML Binary Key | Home | Being Black in America »
Wednesday, 19 October 2005 20:15:40 (GMT Daylight Time, UTC+01:00)
*Highly* unlikely, I'm sure, but just so that we don't get accused of 'lifting' your chosen logo's newspaper graphic further on down the road.

http://www.zeepe.com/zeepeinfo/zeepeapps/imunch/

Jerry
Jerry Mead
Wednesday, 19 October 2005 23:24:46 (GMT Daylight Time, UTC+01:00)
Im wondering what the requirement is for having 3 different subscription types that the user has to choose from. There is no reason (other than time, maybe) that this can't be done as follows.

1. New Subscription
2. Ask user for Uri (or guess it from their clipboard and get them to confirm)
3. Interrogate Uri (including any analysis of what they punched into the Uri field) to see what is actually on the other end
4. Display the appropriete extra customization steps (or ask for clarification if you can't figure out what the other end is)

Keep it as simple as possible for the end user?
Mark Allanson
Wednesday, 19 October 2005 23:47:52 (GMT Daylight Time, UTC+01:00)
Mark,
1. What URI scheme would you suggest for newsgroups that would be familiar to 90% of the population? Our original design used URIs and it was too confusing for me to get right all the time.
2. Subscribing to a search result (e.g. every time "dare obasanjo" shows up in a blog) is often a different flow from wanting to subscribe to a web site, (e.g. wanting to subscribe to Robert Scoble's blog). I can't think of a heuristic for telling one from the other that wouldn't be buggy.

Bottom Line: I don't think having users type in URIs and then having unexpected magic happen is making things more simple. Sometimes being explicit makes things easier.
Dare Obasanjo
Thursday, 20 October 2005 07:53:27 (GMT Daylight Time, UTC+01:00)
I've been interested in the idea / possibility of subscribing to an OPML feed. Let's say, for instance, that I want to add the CodeBetter OPML to my feed list. There's no way for me to know when a new feed joins their OPML, since the OPML import is a one time thing even though the OPML is a dynamic resource (http://codebetter.com/blogs/Opml.aspx?GroupID=2).

I know the standard answer to this is "why would you do that? why not just subscribe to the aggregated feed?". The problem with that is that many aggregated feeds only return the most recent X feeds, so I'd miss posts. There are other problems, but that's the main one. My solution has been to import the OPML, read it for a while, make sure I'm up to date on all posts, delete the folder, and reimport the OPML. I'd love it if RSSBandit did this transparently. I don't know, though, maybe no one else cares about missing a post or two on an aggregated blog site.

Does an OPML subscription seem useful and / or feasible?
Jon Galloway
Thursday, 20 October 2005 12:31:51 (GMT Daylight Time, UTC+01:00)
Hi Dare,

I didn't manage to catch up with you and talk at the MVP summit, but I wanted to let you know that I've been running the Alpha for a while and its been fantastically stable and the CPU problems I was having with the earlier release were not present. Good job man!
Mitch Denny
Thursday, 20 October 2005 18:47:39 (GMT Daylight Time, UTC+01:00)
I agree with the idea of deciding subscription type based on URL.

If I input example.com, it should autodiscover both feeds and opensearch description documents (the latter will be better explained shortly). If there's an OSD, then there should be a textbox where I can input a query.
Michael Fagan
Thursday, 20 October 2005 18:50:54 (GMT Daylight Time, UTC+01:00)
Actually I see now you're using {0}. I don't really like that method... if you're going to support 'query syntax' directly (I prefer supporting the OpenSearch Description Document instead), then why not use the Query Syntax spec http://opensearch.a9.com/spec/opensearchquerysyntax/1.1/ ?
Michael Fagan
Thursday, 20 October 2005 19:40:32 (GMT Daylight Time, UTC+01:00)
Michael,
>I agree with the idea of deciding subscription type based on URL.

You should have read my questions in response to his suggestion. What URL should users use if they want to subscribe to a newsgroup? How many people know the correct syntax for the news or nntp URI schemes? How does guessing the subscription type from URL work when the user wants to create a subscription to every occurene of their name on Technorati or Feedster without asking them to go ahead and write the search via the Web first?

>Actually I see now you're using {0}. I don't really like that method... if you're going to support 'query syntax' directly (I prefer supporting the OpenSearch Description Document instead), then why not use the Query Syntax spec http://opensearch.a9.com/spec/opensearchquerysyntax/1.1/ ?

Maybe it's because we've been using the {0} query syntax since before OpenSearch existed?
Dare Obasanjo
Thursday, 20 October 2005 19:54:14 (GMT Daylight Time, UTC+01:00)
Oh, sorry I missed your newsgroup comment.

Seems to me that newsgroup names can be determined via a regex. Also you can extract those names from news://, google groups urls, and MSR Netscan URLs such as http://netscan.research.microsoft.com/Static/reportcard.asp?searchfor=uw.forsale

Ah, I didn't know you'd be using {0} already. In what other capacity is it used in RSS Bandit?

> How does guessing the subscription type from URL work when the user wants to create a subscription to every occurene of their name on Technorati or Feedster without asking them to go ahead and write the search via the Web first?

I don't quite understand the question..
The user enters 'feedster.com', and it should autodiscover the opensearch description from that.
Michael Fagan
Thursday, 20 October 2005 20:21:36 (GMT Daylight Time, UTC+01:00)
Michael,
It is more straightforward for an end user to know that to subscribe to newsgroups they go hit [New] in the toolbar and select "Newsgroup" from the drop down menu than having them guess how we use regular expressions to guess that they want to subscribe to a newsgroup. Over the past couple of years working on RSS Bandit, I've redesigned and reworked a number that seemed intuitive to me that our users didn't get.

>Ah, I didn't know you'd be using {0} already. In what other capacity is it used in RSS Bandit?

We use it for the feature described in my post. Adding a Web or RSS search engine to RSS Bandit. Of course, most people don't know this feature exists since it is exposed in an obscure manner. Which is one of the reasons it is now part of the subscription wizard.

>The user enters 'feedster.com', and it should autodiscover the opensearch description from that.

You're kidding right? So user's have to know that if they type in the name of a search engine, that sometimes BUT NOT ALWAYS that this would then initiate the process of subscribing to the results of a particular search query. I can just see the bug reports piling in, "How when I type http://search.msn.com I get the subscription wizard but if I type http://www.google.com, it doesn't? Is it because you work for Microsoft?".

Thanks for the suggestion but I'll just keep my existing design if you don't mind.

RSS Bandit has tens of thousands of users, many of them aren't English speakers nor are they familiar with some of the geeky crap folks like you and me take for Bandit.
Dare Obasanjo
Thursday, 20 October 2005 23:21:05 (GMT Daylight Time, UTC+01:00)
Hey, sorry, I don't think I expressed myself that clearly :-) . By all means keep your current design, I wouldn't suggest otherwise.

I'm not proposing removing or replacing any existing functionality, merely adding new functionality. I absolutely agree with the need to make this easier for non-geeks, which is the real point I'm trying to make.

*IF* a url entered in the wizard for adding a feed OR a search isn't a feed, then that's where autodiscovery takes place, correct? That's all I'm really saying. I think that if autodiscovery takes place, and an OSD is found, it can add it to the search wizard, and let the user know about it in some manner.

Syndication searching is still in its early days, methinks.
Michael Fagan
Thursday, 20 October 2005 23:31:45 (GMT Daylight Time, UTC+01:00)
Oh, and in case I didn't make this clear either, I think that having those three options under 'new' is exactly the right way to do it.
Michael Fagan
Comments are closed.