Saturday Aug 29, 2009

Are blogs losing their infuence?

Richard Morgan sent me this article, "Are Blogs Losing Their Authority To The Statusphere?" dated March 10th 2009, which argues that while blog authority ranking according to Technorati remains fairly static, the scores of the various blogs are declining. Technorati uses an inlist scoring algorithm which may be part of the problem, but it would seem to me that micro-blogging is impacting the strength of the voice of blogs as a communications tool, which is what the article argued. In some way's not just micro-blogging, but the various places where people can and do record what they do and think. When I started this blog 5 years ago, I chose to restrain what I put here but other media have grown in popularity, and so people's ability to express themselves have grown. There is a diversification of publication sites which makes following people harder, although technorati only set out to capture blogs, not people, blogs seem no longer to be at the centre of how the internet records what people think. I know that I have been writing less frequently.

Internet messaging is built on a growing distributed architecture, consisting of

  • publication,

  • distribution,

  • aggregation and

  • consumption.

Different sites and technologies seek to perform and excel in different parts of the chain. The aggregation stage permits people to view people, if they permit it, or subject matter and most importantly control their own entry points to the mess that is today's content, by which I mean choose to follow people of look for specific expertise. I think that authors should seek to co-operate with this consumer control of the reading process. It should be noted that the behaviour of individuals and corporations will differ. In particular most media companies want to capture the reader/viewer but individuals have no need to copy this behaviour. I try to post content and let people find it; I hope I have developed a reputation for expertise in some subjects over my career.

By keeping the architecture in mind, one can try and avoid annoying your readers, who, if one has any, are likely to be your friends. Bad habits I see are people who syndicate their tweets into facebook, so I, and others, get to know about their breakfast twice, and I am not a fan of syndicating one's feed into blogs using the APIs. This latter habit annoys me because I don't see the blog as an aggregation tool, but a publishing tool, and so I expect original work, of some description in people's blogs. This can be even worse when people then publicise the blog, containing bookmarks using a micro blog. That's three clicks to read something written by someone other than the person who's views you've subscribed to, and if using a wireless device that's a real pain. NB This is also true if you subscribe to Digg feeds, you get to 'interesting' content via the Digg page, so three clicks, three tabs or windows to read content you want. Another offence which I wish I could deal with more easily is the microblogging incontinent. The only way I have discovered how to deal with those, is to unsubscribe.

One can, and I do aggregate my feeds into one place. I originally created a personal planet, which aggregates some of the feeds I create. I have tried to create an everything feed at, which also has a nice key of the feeds I contribute to. This means that my readers can construct a feed that interests them. I know that some friends are interested in the technology, but not the politics. I commit the offence of subscribing my friend feed to face book, but I consider Facebook to be mainly a consumer. I need to think about this. Its not great, but I don't syndicate my tweets directly to my face book statuses (sic), nor do I copy them back into friend feed. Manging my facebook feed is not easy and is compounded by Facebook's desire to perform all roles in the architecture while being 'open'. Its this open-ness which has enable site specialisation around, for instance, travel, books, restaurants and even at living social, iphone apps.

I suppose I am appealing for people to consider what tools they use to perform a specific role in the the personal content architecture. Don't over aggregate, if people are interested in your thoughts they'll find them. Don't shove it down their throats.

When I first considered writing this little essay, it seemed interesting to consider, “Is the status-sphere replacing blogs?”, others including Tim Bray have written about this since and argue Not. I hope that the evolution of easy micro-blogging, will free blogs to become deeper and more interesting. I know that I have produced less frequent blog article since I took up with Twitter, but I also considered my feed, to be a microblog of sorts. Another key development is that the use of sites like has turned in-list search ranking from a vote of web authors, where you needed the technology skills and resource to have a web page in order to influence the sort order, into a vote of web readers and authors. The ease of micro blog adoption means that an even large crowd should now be participating in the construction of in list search orders. I am unsure how url shortner's impact the search engines in-list calculations. They make it harder, I 'm sure, as does the fact there's more than one. Many argue that Twitter's best value is as a search engine and that it, and other micro-blogging systems won't replace blogs because there are too many subjects that can't be accurately discussed in 140, characters. Techcrunch published further thoughts on twitter, and it chances of supplanting the blogs, however it takes less time to tweet, rubbish gets lost easier, and twitter in particular is designed to be used by handheld devices. (I don't think I'd like to have written this on my new Nokia 5800, or even my ipodtouch.) It should be noted, that while its very easy to create a 140 character message, it should be easy to create a podcast or even a video, but its not. They are both difficult to create, especially if you don't just record a chat amongst friends but try and 'deliver/perform' a report. This is a skills issue. They wouldn't pay Steven Fry all that money to make audio books if it was easy.

One final thought is that communitarian aggregation is not well done at the moment. One of the strength's of Peter Reiser's approach, 'Community Equity', to knowledge management systems that at its heart is a personal rating engine. (See also They don't yet have this as truly n-dimensional, which I think is needed, so you can rate your own content, rate other author's expertise, rate & describe their interest to you. I may play with a Google App or Zembly, to experiment with some of how to make some of this work. A very rich inter-personal network with sophisitcated popular and machine calculated relevance scoring is something that can add value. Content could flow through your colleagues votes moving closer and further away from your viewing window and your friends and colleagues advice and hints would influence or determine what you find. Google reader's share facility is quite close, but there's only one dimension, you can have friends, and they can recommend stuff for you to read. (I think more can be done.)

I hint that one of Technorati's problems is its reliance on in-list. Searching the Workplace Web, written by Fagin, Kumar and McCurley, which I commented on, on this blog in an article called, The Shape of Internet, write about a number of relevance and ordering tests that could be used and specifically argue that within the corporate intranet other sorts and relevance tests may be more appropriate to help solve a number of questions such as authority. They also argue that intranet URL naming is less search friendly and it is clear that the dissipation of people's voice and advice over multiple sites with different naming conventions, often using surrogates or numbers and URL shortners means that the internet is catching up on the early intranet in the chaos of name space. It may be time to review in-list and begin to weight votes for relevance and sort-order.

Are blogs losing their relevance, maybe, maybe not. Well written opinions by disinterested experts will always be valued. As the dross moves to the microblogs, this may liberate the blogs to re-establish themselves as clear voices of expertise. Some of what was observed Richard's post to me may be failures in Technorati, its initial insights are aged and its being out innovated.


Thursday Jul 02, 2009


I have reduced the number of tags available in the banner. You can still use Google, or the Yesterdays Words page here which has a number of search tools for this blog.

Thursday Apr 02, 2009

A second look at Second Brain

I have revisited secondbrain recently and decided I need to get to grips with its libraries and collections. I am not sure of the differences and whether I should be create broad large collections such as travel or software, or even something narrower but broad such as database, or use it for more project orientated collections such as specific journies or personal engineering tasks, a bit like what it takes to justify a new snipsnap page on my bliki. I quite like the fact they give me a domain name, and that I got their early enough to get "davelevy".

It now takes a much broader range of feeds, which was the criticism I made last time I reviewed it, and creates an aggregated tag cloud. This is neat, but I hadn't realised how many tags my picture collection generates. The tag cloud is dominated by the places tag and the geographic qualifiers. When you add the bookmarks created while planning the travel, it dominates the tag cloud, which I am not sure is what I want. (I wonder if they could or should permit us to weight the tags by feed.) Usability is also inhibited here because like most people, I don't tag a feed as belonging to itself, so my bookmarks aren't tagged as bookmarks. Also several of my feeds are not tagged at all. All-in-all, this is a feature I like, so I'd really like a tag cloud widget.

I was looking at second brain to see if I could make it my home page and consolidate the various sites I am using into one place, it could well be possible. I'd loose control of my look and feel, and I'd need to consider how to host original textual content but a blog might work for that if I have SB's collections and libraries. Perhaps I'll try and migrate one of my travel pages to SB and see what it looks like, and how useful I find it.

tags: ""

Monday Sep 01, 2008

wordle clouds

I was browsing sun blogs, when I was pointed at this blog, "Startups" which has an article on wordle. This produces word maps from feed URLs. You can see a view of mine below, it was taken earlier today.

wordle word map

Wordle is (cc) Attribution only, so I need to say the picture above is produced by [site | faq ]. Its a Java App so sadly, I had to screenscrape the full size picture, and so generating one per article is too much of a fag, and hard to do ; they use the feed and so see many articles. Word mapping is becoming a powerful tool, but this needs to be licensed by the blog publisher to be included as part of the blog server functionality. I think we'd want clouds/article and clouds/blog. I thought I'd check out what happens if one uses a feed, so I also did one on See below. It suggests to me that wordle has a fairly short view back in history.

wordle of

Of course both my blog and the sun bloggers site have hyperlinked tag clouds, which again diminished the use of wordle to us. I wonder if it teaches us anything for other applications such as our "CEC Messaging Platform".

tags: ""

Wednesday Apr 23, 2008


In the article My Twitter Manifesto, I stated that I did not propose to use Twitter as a microblog, the closest thing I have to one is my feed, which since I bookmark nearly everything I read, gives you an idea of what I am thinking about at least. I nearly always comment on a site/page, although not always in my own words. But if your interested.... it's there. The most recent seven items are also in the sidebar on this page.


Wednesday Jan 30, 2008

Mashing up with Snipsnap

Having run a personal snipsnap wiki for a number of months, I have got to the stage where I want to use it as a host for a couple of mashups. Fortunately there is a user contributed macro that allows for the use of HTML within a snipsnap page. Otherwise, because the page input screen interprets a markup language, all HTML is treated as text. This macro is installed by copying it i.e. the .jar file to the directory named in the Java CLASSPATH. It really is as simple as that, and then using snipsnap's macro syntax. You can also see what I have written at my bliki.

<SPAN STYLE="color:red">This font is Red caused by a Span tag pair</SPAN>

This has allowed me to embed a google map in one of my pages, the example below is about Brussels.


Snipsnap + Google Map


and the example below shows a tag roll.


Snipsnap + tag roll


In both cases, I have implemented then using <IFRAME> tags; in Google's case because they suggest it and in the case because it is a Javascript and the CSS wasn't working. It seemed easier to escape the CSS domain by using IFRAME as I have had difficulty with this problem before. I think its to do with the anchor rules I define in the CSS files. I have had this problem with roller as well, although not with the tag cloud as you can see my checking out My Links page at, although it remains marked as beta. You can also see what I have written at my bliki, but again, its not much more.


Thursday Nov 30, 2006

Tags in Roller & Technorati

Two days ago, I looked at how to display the tags assigned to an article hosted in roller and left the question of generating techorati tags unfinished.

<!-- Now lets technorati tag the tags -->
<P><small>technorati tags:
    #set ($mytags=$entry.tags)
    #foreach ( $tag in $mytags )
        <a rel="tag" href="$">$</a>
        <!-- NB there is a trailing space on the line above #end -->

should produce the technorati tag list in a small font paragraph. Now on to Chris's challange.  

I have a challenge about how to migrate forward, since if I put this code in the _day file (inside the entries loop), than I will get the technorati tags twice, for those articles with both roller tags and technorati tags specified. This is caused by my eagerness to load my tags into roller to see my tag cloud; thats why I have articles with both roller and techorati tags. I might be able to create a local velocity macro and use this as content within an article, but I think the _day file is the right place. Have to think about it.

The tags are thus still done by hand.



Monday Nov 27, 2006

Tuning tags for query & propagation

I am still hand coding the the tags in this blog. Roller has a macro to display a hyperlinked tag list but resolves the links inside the roller instance.

<P><small>tags: #showEntryTags($entry)</small></P>

However, I still want to use technorati, so I am copying the tags from the input text box into the content body and using the topic plugin to create my technorati tags. (Actually I don't even necessarily always keep the same tags, but I am pretty consistent.) I wonder how to write velocity code to build the technorati tag list.


Thursday Nov 02, 2006

More about Roller tagging

I have just been experimenting with the tag macros. It is possible to display an articles roller tags using the #showEntryTags($entry) macro. This can be inserted in the entries loop in _day file. 

Because I am planning to continue to use the technorati tags, unless I plan to use a hidden display format I would have two copies of the tag list in the HTML rendered version. So I have not implemented this macro in the _day file. 

I am struggling to get the tag cloud displayed, it seems to refuse to interpret the macro call

The #showTagCloud macro was not implemented since the velociraptor code is so simple (yeah right, I couldn't work it out). I now have a tag cloud implemented on the "About Me" page; Dave Johnson gave me some sample code to make it work. There were other reasons for not implementing showTagCloud as a macro. The team were unclear given the number of customisations that would be required, how used the macro would be, because no matter how few used it, it would need to be maintained.

First set up some CSS rules, roller offers five levels of weight, so five rules is good. 

          .s1 {font-size:60%;}
          .s2 {font-size:80%;}
          .s3 {font-size:100%;}
          .s4 {font-size:120%;}
          .s5 {font-size:140%;}

These are relative rules, which is jolly good so they inherit from the Body or Paragraph rules, but you can make the rules as complex as you want. The following velociraptor/roller code implements the tag cloud in the html file.

          #set($mytags = $model.weblog.getPopularTags(-1, 100))
          #foreach ($tag in $mytags)
              <a class="tag s${tag.intensity}" href="$url.tag($"

Note the way that the CSS rule names agree with the class attribute of the anchor tag.

This article was updated at 13:30 GMT 3rd November GMT.


Monday Aug 28, 2006

Semantic footprints through time

Another thought (about tags & prompting, see also here... or here...) is that prompting (& categories) inhibits evolution. Either the ideas being explored change, or the common use of words chnage. An example on this blog is that I should probably create a travel category on this blog rather than treat it as a qualifier of Culture, but because I am prompted, I have never bothered.

On the other hand, I also want you to able to find all the articles in a natural series, so I value the stability but its means that you have to learn about my use of the catagories. The search engine and article list on the sidebar are also category context sensitive which is good.

(I really need to see the coming Roller tags implementation, to see if tags and categories really are interchangable).


Tuesday May 30, 2006

Visualising tag clouds

Two points come to mind on reflecting my conversation with Elias Torres at www20006, (See Tags and Spontaneity below...). First maybe on tag clouds we should use colour for highly used & less frequently used with Red being highly used and Blue (or Indigo) less frequently used. This should mean that the less frequently used tags, which are the most discriminatory (i.e. meaningful) are not visually eclipsed by the most heavily used.

Secondly, if we were to look at the structure "UNIX > Solaris > AIX, the significance would only be true if all taggers tagged the articles as UNIX & Solaris or UNIX & AIX, and only then could we be clear on the meaningfulness of a tag. This illustrates that a stranger may need to be familar with the crowd's use of language.

Thirdly I'm also not sure how we might 'refine' a query if we start with very meaningful tags, we would have to re-query, although the interface offers you a list of associated tags even for the smallest of queries.


Tuesday May 23, 2006

Tags and Spontaneity

We broke for coffee and I bumped into Elias Torres, one of the roller development team, so we had a long chat about tags on the web , looked at how to make roller's implementation better and how I could become more helpful by creating an environment to test this stuff. He pointed out that my Qube (currently not working) needs replacing.

We tried to remember the state of our debate as we left it. Elias co-authored a paper "Ranking Bookmarks & Bistros", one of the conclusions is that tagging software should not be too helpful in prompting for tags; the spontaneity of the tag author is an important part of the folksonomy. My view is that a tagging correspondent will have a personal structure to the tags they use, in my case I have a small group of 1st order tags, which will always be used, or nearly always. I believe that my contribution to the folksonomy is more effective if I am helped to keep to my structure. For my medium order tags, I want to be prompted as to my spelling and which words I use to imply meaning. Elias' argument is that the collective expression is stronger if more spontaneous (I think). I think that part of our disagreement is that I focus on tagging for personal use and at time of discovery and hence initial tagging, Elias is concerned about the stranger looking for wisdom in the tag base.

For instance for my retrievals I like to go in large or medium and refine my queries using the "+" function adding filters to my query until the page is small enough to inspect. (Is this a binary chop?) It all reminds me that I need to ask him for the data model (author, articles, articles, tags) numbers and growth rates. I wonder if Linda or Will can give them to me from


Sunday Jan 15, 2006

Follow my steps at (, or here)

Thanks to Jim Grisanzio who rendered his feed on his blog, using their link roll gadget. (You may need to log into to make it work).

I have implemented mine here in my sidebar panel, with a limit of 7, so you'll need to come to my HTML version of the page to see it. I need to work on the HTML style that the roll is published in, but I shall be amending my personal sites to use both this and the tags tool. Because I bookmark a lot of what I browse, by reading my feed (the hyperlink is in the sidebar), you can get some idea of what I am (or am planning) to do.

I was busy doing some DIY this weekend, so the new links at the moment are not based on the weekend's reading, which my monday links often are. I am considering assembling a computer of my own, which for thos of you who know me is a big step, so we have a few e-shopping sites that will sell me stuff, although these are not recommendations. I'm merely sharing my research.


Tuesday Jan 03, 2006

Implementing Tags in a Database??

I was considering why doesn't offer a cloud for tag based queries, while it does offer the cloud for User based queries, and I received the reply that it was (probably) related to query cost, which set me thinking. What would the database look like?

 entity model

Is this right, look at the para below, the requirement that a tag is used by someone to describe the bookmark is lost in above version of the model.

Surely we have only three entities, User, Tag & Bookmark (also known as URL). Both User & Tag have a many to many relationship with each Bookmark. We can resolve these many2manies in two ways, by using an allocation table, or by adopting a meta model model.

The data model must take into account the fact that a User describes a bookmark using one or more tags. i.e. DaveLevy describes as (UK web World News politics). Other people have used different tags, although only a few associated it with journalism. Each of these user, bookmark, tag relationships must be stored seperately; otherwise we loose the user's relationship and ownership of the tag set. This implies/mandates that somwhere a user/bookmark/tag database object (either table or index is required). This means that the bookmark entity must be related to the user by an allocation. This is required to implement the many-to-many, but also to allow the tags to be owned by a user. The tag must also have an allocation between the Bookmark & itself but is owned by the defining user i.e. again we must have a User/Bookmark/Tag intersection entity. The model immediately below meets this requirement, although implementing the Tag attribute as a membership of the tag - user/bookmark relationship and implementing it as a foreign key is also a possibility.

Dave's refined ERD

I have not documented the name of the relationship between tag, bookmark and their allocations. This becomes quite hard because we have transformed the entity into "operational masters" and they are likely to be implemented as indexes. The difficulty in naming the relationship between these entites and the allocation entities implies that we have modelled the problem well. Neither does the diagram above illustrate mandatory/optional attributes of the relationships. The Knows of relationship is optional. During the period between registration and the first bookmark, a User will have zero user/bookmark allocations. I suppose it is possible to enter a Bookmark without tagging it, which makes the Described by relationship optional as well. If either of the allocations exist though, their masters must exist.

It should be noted that any domain of definition can be applied to the bookmark entity (or at least its key). By stating that the bookmark must be (say) a roller article URL, we have a viable tag model for roller articles. In fact, I have considered opening a new account exclusively to act as a blog index and to provide a tag/cloud map for this blog.

I am next going to look at some queries and the relational algebra that can be applied to this model; the reason I developed the model was to examine the performance implications of different entry points.

tags: ""




« June 2016