Ivan’s private site

January 18, 2014

Some W3C Documents in EPUB3

Filed under: Code,Digital Publishing,Python,Work Related — Ivan Herman @ 13:04
Tags: , ,

I have been having fun the past few months, when I had some time, with a tool to convert official W3C publications (primarily Recommendations) into EPUB3. Apart from the fact that this helped me to dive into some details of the EPUB3 Specification, I think the result might actually be useful. Indeed, it often happens that a W3C Recommendation consists, in fact, of several different publications. This means that just archiving one single file is not enough if, for example, you want to have those documents off line. On the other hand, EPUB3 is perfect for this; one creates an eBook contains all constituent publications as “chapters”. Yep, EPUB3 as complex archiving tool:-)

The Python tool (which is available in github) has now reached a fairly stable state, and it works well for documents that have been produced by Robin Berjon’s great respec tool. I have generated, and put up on the Web, two books for now:

  1. RDFa 1.1, a Recommendation that was published last August (in fact, there was an earlier version of an RDFa 1.1. EPUB book, but that was done largely manually; this one is much better).
  2. JSON-LD, a Recommendation published this week (i.e., 16th of January).

(Needless to say, these books have no formal standing; the authoritative versions are the official documents published as a W3C Technical Report.)

There is also draft version for a much larger book on RDF1.1, consisting of all the RDF 1.1 specifications to come, including all the various serializations (including RDFa and JSON-LD). I say “draft”, because those documents are not yet final (i.e., not yet Recommendations); a final version (with, for example, all the cross-links properly set) will be at that URI when RDF 1.1 becomes a Recommendations (probably in February).

March 1, 2013

RDFa 1.1, microdata, and turtle-in-HTML now in the core distribution of RDFLib

This has been in the works for a while, but it is done now: the latest (3.4.0 version) of the python RDFLib library has just been released, and it includes and RDFa 1.1, microdata, and turtle-in-HTML parser. In other words, the user can add structured data to an HTML file, and that will be parsed into RDF and added to an RDFLib Graph structure. This is a significant step, and thanks to Gunnar Aastrand Grimnes, who helped me adding those parsers into the main distribution.

I have written a blog last summer on some of the technical details of those parsers; although there has been updates since then, essentially following the minor changes that the RDFa Working has defined for RDFa, as well as changes/updates on the microdata->RDF algorithm, the general approach described in that blog remains valid, and it is not necessary to repeat it here. For further details on these different formats, some of the useful links are:


November 6, 2012

RDFa 1.1 and microdata now part of the main branch of RDFLib

Filed under: Code,Python,Semantic Web,Work Related — Ivan Herman @ 21:34
Tags: , , , ,

A while ago I wrote of the fact that I have adapted my RDFa and microdata to RDFlib. Although I did some work on it since then, nothing really spectacular happened (e.g., I have updated the microdata part to the latest version of the microdata->RDF conversion note, and I have also gone through the tedious exercise to make the modules usable for Python3).

Nevertheless, a significant milestone has been reached now, but this was not done by me but rather by Gunnar Aastrand Grimnes, who “maintains” RDFlib: the separate branch for RDFa and microdata has now been merged with the master branch of RDFLib on github. So here we are; whenever the next official release of RDFLib comes, these parsers will be part of it…

August 31, 2012

RDFa, microdata, turtle-in-HTML, and RDFLib

For those of us programming in Python, RDFLib is certainly one of the RDF packages of choice. Several years ago, when I developed a distiller for RDFa 1.0, some good souls picked the code up and added it to RDFLib as one of the parser formats. However, years have gone by, and have seen the development of RDFa 1.1, of microdata, and also the specification of directly embedding Turtle into HTML. It is time to bring all these into RDFLib…

Some times ago I have developed both a new version of the RDFa distiller, adapted for the  1.1 RDFa standard, as well as a microdata to RDF distiller, based on the Interest Group note on converting microdata to RDF. Both of these were packages and applications on top of RDFLib. Which is fine because they can be used with the deployed RDFLib installations out there. But, ideally, these should be retrofitted into the core of RDFLib; I have used the last few quiet days of the vacation period in August to do just that (thanks to Niklas Lindström and Gunnar Grimes for some email discussion and helping me through the hooplas of RDFLib-on-github). The results are in a separate branch of the RDFLib github repository, under the name structured_data_parsers. Using these parsers here is what one can do:

g = Graph()
# parse an SVG+RDF 1.1 file an store the results in 'g':
g.parse(URI_of_SVG_file, format="rdfa1.1") 
# parse an HTML+microdata file an store the results in 'g':
g.parse(URI_of_HTML_file, format="microdata")
# parse an HTML file for any structured conent an store the results in 'g':
g.parse(URI_of_HTML_file, format="html")

The third option is interesting (thanks to Dan Brickley who suggested it): this will parse an HTML file for any structured data, let that be in microdata, RDFa 1.1, or in Turtle embedded in a <script type="text/turtle">...</script> tag.

The core of the RDFa 1.1 has gone through a very thorough testing, using the extensive test suite on rdfa.info. This is less true for microdata, because there is not yet an extensive test suite for that one yet (but the code is also simpler). On the other hand, any restructuring like that may introduce some extra bugs. I would very much appreciate if interested geeks in the community could install and test it, and forward me the bugs that are still undeniably there… Note that the microdata->RDF mapping specification may still undergo some changes in the coming few weeks/months (primarily catching up with some development around schema.org); I hope to adapt the code to the changes quickly.

I have also made some arbitrary decisions here, which are minor, but arbitrary nevertheless. Any feedback on those is welcome:

  • I decided not to remove the old, 1.0 parser from this branch. Although the new version of the RDFa 1.1 parser can switch into 1.0 mode if the necessary switches are in the code (e.g., @version or a RDFa 1.0 specific DTD), in the absence of those 1.1 will be used. As, unfortunately, 1.1 is not 100% backward compatible with 1.0, this may create some issues with deployed applications. This also means that the format="rdfa" argument will refer to 1.0 and not to 1.1. Am I too cautious here?
  • The format argument in parse can also hold media types. Some of those are fairly obvious: e.g., application/svg+xml will map on the new parser with RDFa 1.1, for example. But what should be the default mapping for text/html? At present, it maps to the “universal” extractor (i.e., extracting everything).

Of course, at some point, this branch will be merged with the main branch of RDFLib meaning that, eventually, this will be part of the core distribution. I cannot say at this point when this will happen, I am not involved in the day-to-day management of the RDFLib development.

I hope this will be useful…

April 24, 2012

Moved my RDFa/microdata python modules to github

Filed under: Code,Python,Work Related — Ivan Herman @ 12:37
Tags: , , , ,

In case you were using/downloading my python module for RDFa 1.1 or for Microdata: I have now moved away from the old CVS repository site, and moved to GitHub. The two modules are on the RDFLib/pyrdfa3 and RDFLib/pymicrodata repositories, respectively. Both of these modules are more or less final (there are still some testings happening for RDFa, but not much left) and I am just happy if others chime in in the future of these modules.

Although part of the RDFLib project on GitHub, the two modules are pretty much independent of the core RDFLib library, although they are built on top of it. I hope that, with the help of people who know the RDFLib internal structures better, both modules can become, eventually, part of the core. But this may take some time…

December 16, 2011

Where we are with RDFa 1.1?

English: RDFa Content Editor

Image via Wikipedia

There has been a flurry of activities around RDFa 1.1 in the past few months. Although a number of blogs and news items have been published on the changes, all those have become “officialized” only the past few days with the publication of the latest drafts, as well as with the publication of RDFa 1.1 Lite. It may be worth looking back at the past few months to have a clearer idea on what happened. I make references to a number of other blogs that were published in the past few months; the interested readers should consult those for details.

The latest official drafts for RDFa 1.1 were published in Spring 2011. However, lot has happened since. First of all, the RDFWA Working Group, working on this specification, has received a significant amount of comments. Some of those were rooted in implementations and the difficulties encountered therein; some came from potential authors who asked for further simplifications. Also, the announcement of schema.org had an important effect: indeed, this initiative drew attention on the importance of structured data in Web pages, which also raised further questions on the usability of RDFa for that usage pattern This came to the fore even more forcefully at the workshop organized by the stakeholders of schema.org in Mountain View. A new task force on the relationships of RDFa and microdata has been set up at W3C; beyond looking at the relationship of these two syntaxes, that task force also raised a number of issues on RDFa 1.1. These issues have been, by and large, accepted and handled by the Working Group (and reflected in the new drafts).

What does this mean for the new drafts? The bottom line: there have been some fundamental changes in RDFa 1.1. For example, profiles, introduced in earlier releases of RDFa 1.1, have been removed due to implementation challenges; however, management of vocabularies have acquired an optional feature that helps vocabulary authors to “bind” their vocabularies to other vocabularies, without introducing an extra burden on authors (see another blog for more details). Another long-standing issue was whether RDFa should include a syntax for ordered lists; this has been done now (see the same blog for further details).

A more recent important change concerns the usage of @property and @rel. Although usage of these attributes for RDF savy authors was never a real problem (the former is for the creation of literal objects, whereas the latter is for URI references), they have proven to be a major obstacle for ‘lambda’ HTML authors. This issue came up quite forcefully at the schema.org workshop in Mountain View, too. After a long technical discussion in the group, the new version reduces the usage difference between the two significantly. Essentially, if, on the same element, @property is present together with, say, @href or @resource, and @rel or @rev is not present, a URI reference is generated as an object of the triple. I.e., when used on a, say, <link> or <a> element, @property  behaves exactly like @rel. It turns out that this usage pattern is so widespread that it covers most of the important use cases for authors. The new version of the RDFa 1.1 Primer (as well as the RDFa 1.1 Core, actually) has a number of examples that show these. There are also some other changes related to the behaviour of @typeof in relations to @property; please consult the specification for these.

The publication of RDFa 1.1 Lite was also a very important step. This defines a “sub-set” of the RDFa attributes that can serve as a guideline for HTML authors to express simple structured data in HTML without bothering about more complex features. This is the subset of RDFa that schema.org will “accept”,  as an alternative to the microdata, as a possible syntax for schema.org vocabularies. (There are some examples on how some schema.org example look like in RDFa 1.1 Lite on a different blog.) In some sense, RDFa 1.1 Lite can be considered like the equivalent of microdata, except that it leaves the door open for more complex vocabulary usage, mixture with different vocabularies, etc. (The HTML Task Force will publish soon a more detailed comparison of the different syntaxes.)

So here is, roughly, where we are today. The recent publications by the W3C RDFWA Working Group have, as I said, ”officialized” all the changes that were discussed since spring. The group decided not to publish a Last Call Working Draft, because the last few weeks’ of work on the HTML Task Force may reveal some new requirements; if not, the last round of publications will follow soon.

And what about implementations? Well, my “shadow” implementation of the RDFa distiller (which also includes a separate “validator” service) incorporates all the latest changes. I also added a new feature a few weeks ago, namely the possibility to serialize the output in JSON-LD (although this has become outdated a few days ago, due to some changes in JSON-LD…). I am not sure of the exact status of Gregg Kellogg’s RDF Distiller, but, knowing him, it is either already in line with the latest drafts or it is only a matter of a few days to be so. And there are surely more around that I do not know about.

This last series of publications have provided a nice closure for a busy RDFa year. I guess the only thing now is to wish everyone a Merry Christmas, a peaceful and happy Hanukkah, or other festivities you honor at this time of the year.  In any case, a very happy New Year!

Enhanced by Zemanta

November 23, 2010

My first mapping from RDB to RDF using a direct mapping, cont.

A few days ago I posted a blog on how the RDB to RDF direct mapping could be used for a simple example. I do not want to repeat the whole blog: the essence of it was that database tables were mapped onto a simple RDF Graph (this is what the direct mapping does) and the resulting graph was transformed into the “target” graph using the following SPARQL 1.1 construct:

  ?id a:title ?title ;
    a:year  ?year ;
    a:author _:x .
  _:x a:name ?name ;
    a:homepage ?hp .
  SELECT (IRI(fn:concat("http://...",?isbn)) AS ?id)
          ?title ?year ?name
         (IRI(?homepage) AS ?hp)
    ?book a  <Book>;
       ?isbn ;
       ?title ;
        ?year ;
       ?author .
    ?author a  <Author>;
       ?name ;
       ?homepage .

where the trick was to use a nested SELECT whose main job was to create URI references from strings. I realized that if one uses the latest editors’ version of SPARQL 1.1 (i.e., that version that is much closer to what SPARQL 1.1 will be) then the solution is actually simpler due to the variable assigning possibility that makes the nested SELECT unnecessary:

  ?id a:title ?title ;
    a:year  ?year ;
    a:author _:x .
  _:x a:name ?name ;
    a:homepage ?hp .
  ?book a  <Book>;
     ?isbn ;
     ?title ;
      ?year ;
     ?author .
  ?author a  <Author>;
     ?name ;
     ?homepage .
  BIND (IRI(fn:concat("http://...",?isbn)) AS ?id)
  BIND (IRI(?homepage) AS ?hp)

which makes, at least in my view, the mapping even clearer.

But SPARQL is not the only way to transform the graph. Another possibility is to use RIF Core. Essentially the same transformation can indeed be expressed using the RIF Presentation syntax. Here it is (with a little help from Sandro Hawke and Harold Boley):

Forall ?book ?title ?author ?isbn ?year ?id (
  ?id[a:year->?year a:title->?title a:author->?author] :-
      ?book[rdf:type-> <Book>
      External(pred:iri-string(?id External( func:concat("http://..." ?isbn ) )))
Forall ?author ?name ?hp ?homepage (
 ?author[a:name->?name a:homepage->?hp] :-
        ?author[rdf:type-> <Author>
        External(pred:iri-string(?hp ?homepage))

(as I did in the earlier examples, I did not put the prefix declaration and other syntactic stuffs into the code above.)

The only difference between the two is that I retained the URI for the author, because generating a blank node on the fly in RIF Core does not seem to be possible. A better solution would be, probably, to mint a URI from the ?author variable just like I did for the ISBN value. Other than that, the two solutions are pretty much identical…

November 19, 2010

My first mapping from RDB to RDF using a direct mapping

A few weeks ago I wrote a blog on my first RDB to RDF mapping using R2RML; the W3C RDB2RDF Working Group had just published a first public Working Draft for R2RML. That mapping was based on a specific mapping language (i.e., R2RML). R2RML relies on an R2RML processing done by, for example, the database system, interpreting the language, using some SQL constructions, etc. The R2RML processing depends on the specific schema of the database which guides the mapping.

As I already mentioned in that blog, a “direct” mapping was also in preparation by the Working Group; well, the first public Working Draft of that mapping has just been published. That mapping does not depend on the schema of the database: it defines a general mapping of any relational database structure into RDF; only a base URI has to be specified for the database, everything else is generated automatically. The resulting RDF graph is of course much more coarse than the one generated by R2RML; whereas the result of an R2RML mapping may be a graph using well specified vocabularies, for example, this is not the case for the output of the direct mapping. But that is not really a problem: after all, we have SPARQL or RIF to make transformation on graphs! Ie, the two approaches are really complementary.

What I will do in this blog is to show how the very same example as in my previous blog can be handled by a direct mapping. As a reminder: the toy example I use comes from my  generic Semantic Web tutorial. Here is the (toy) table:

which is then converted into an RDF Graph:

(Just as in the previous case I will ignore the part of the graph dealing with the publisher, which has the same structure as the author part. I will also ignore the prefix definitions.)

The direct mapping of the first and second tables is pretty straightforward. The URI-s are a bit ugly but, well, this is what you get when you use a generic solution. So here it is:

@base <http://book.example/> .
<Book/ID=0006511409X#_> a <Book> ;
  <Book#ISBN> "0006511409X" ;
  <Book#Title> "The Glass Palace" ;
  <Book#Year>  "2000" ;
  <Book#Author> <Author/ID=id_xyz#_> .

<Author/ID=id_xyz#_> a <Author> ;
  <Author#ID> "id_xyz" ;
  <Author#Name> "Ghosh, Amitav" ;
  <Author#Homepage> "http://www.amitavghosh.com" .

Simple, isn’t it?

The result is fairly close to what we want, but not exactly. First of all, we want to use different vocabulary terms (like a:name). Also, note that the direct mapping produces literal objects most of the time, except when there is a “jump” from one table to another. Finally, the resulting graph should use a blank node for the author, which is not the case in the generated graph.

Fortunately, we have tools in the Semantic Web domain to transform RDF graphs. RIF is one possible solution; another is SPARQL, using the CONSTRUCT form. Using SPARQL is an attractive solution because, in practice, the output of the direct mapping may not even be materialized; instead, one would expect a SPARQL engine attached to a particular relational database, mapping the SPARQL queries to the table on the fly. I will use SPARQL 1.1 below because that gives nice facilities to generate RDF URI Resources from strings, i.e., to have “bridges” from literals to URI-s. Here is a possible SPARQL 1.1 query/construct that could be used to achieve what we want:

  ?id a:title ?title ;
    a:year  ?year ;
    a:author _:x .
  _:x a:name ?name ;
    a:homepage ?hp .
  SELECT (IRI(fn:concat("http://...",?isbn)) AS ?id)
          ?title ?year ?name
         (IRI(?homepage) AS ?hp)
    ?book a <Book> ;
      <Book#ISBN> ?isbn ;
      <Book#Title> ?title ;
      <Book#Year>  ?year ;
      <Book#Author> ?author .
    ?author a <Author> ;
      <Author#Name> ?name ;
      <Author#Homepage ?homepage .

Note the usage of a nested query; this is used to create new variables representing the URI references to be used by the outer query. The key is the IRI operator. (Both the nesting and the AS in the SELECT are SPARQL 1.1 features.)

That is it. Of course, the question does arise: which one would one use? The direct mapping or R2RML? Apart from the possible restriction that the local database system may implement the direct mapping only, it becomes also a question of taste. The heavy tool in R2RML is, in fact, the embedded SQL query; if one is comfortable with SQL than that is fine. But if the user is more comfortable with Semantic Web tools (e.g., SPARQL or RIF) then the direct mapping might be handier.

(Note that these are evolving documents still. I already know that my previous blog is wrong in the sense that it is not in line with the next version of R2RML. Oh well…)

November 2, 2010

My first mapping from RDB to RDF using R2RML

The W3C RDB2RDF Working Group has just published a first public Working Draft for the standardized RDB->RDF mapping language called R2RML. I decided that the only way to understand a specification like that is to try to use it for an example. Caveat: this is a “First Public Working Draft” for R2RML, so many things still have to happen and there will be changes.

For several years now I use a simple example in my generic Semantic Web tutorial (see, e.g., the one at SemTech). It is an artificial example referring to an imaginary bookshop’s table:

which is then converted into an RDF Graph:

(And the tutorial story is how this graph can be merged with a graph coming from another bookshop’s data.) Up until now I always glossed over how this mapping is done. Well, so how could that be done with R2RML?

R2RML defines mappings that describe how an RDB table is mapped on triples. (R2RML is in itself in RDF, b.t.w.) Simply put, in R2RML, each row of a table is mapped to an RDF subject; the individual cells, with the column names, provide the object and the predicates, respectively.

If we look at the middle table in the example, it corresponds to the lower right hand part of the graph. The R2RML mapping has to specify that the homepage column should actually produce an RDF Resource as a literal and not a string. Furthermore, the first column should become a blank node; that has to be specified, too. Here is the way this is all specified:

:Table2 rdf:type rr:TriplesMap ;
    rr:logicalTable "Select  ("_:" || ID) AS pid, Name, ("<" || Homepage || ">) AS Home from person_table";
    rr:subjectMap [ a rr:BlankNodeMap ; rr:column "pid" ; ] ;
    rr:propertyObjectMap [ rr:property a:name; rr:column "Name" ] ;
    rr:propertyObjectMap [ a rr:IRIMap ; rr:property a:homepage; rr:column "Home" ] .

What happens here is:

  1. a mapping is defined that turns the original table into a virtual, “logical” table using SQL. The goal here is to generate a blank node ID on the fly, and a URI in NTriple syntax (note, however, that I am not sure it is o.k. to use that approach in the spec!);
  2. the subject for the triples is chosen to be a cell in a specific column (“pid”, generated by the SQL transform of the previous point), and it is also specified that this is a blank node;
  3. the other two properties are specified (for the same subject); the one for the home page also specifies that the object must be a URI resource (as opposed to a Literal).

That is it. Mapping of the bottom table to the lower left hand corner of the graph is also quite similar, I will not go into this here.

But we still need the “root”, so to say, i.e., the node in the upper right hand corner, the top portion of the graph (with the title and the year) and, mainly, we also have to relate the root to the portion of the graph that is generated from the middle table.

First, the following R2RML part does the job of generating the top part of the graph:

:Table1 rdf:type rr:TriplesMap ;
    rr:logicalTable "Select ('<http:..isbn/' || ISBN || '>') AS isbn, 
                     Author, Title, Publisher, Year from book_table";
    rr:subjectMap [ rdf:type rr:IRIMap ; rr:column "isbn" ] ;
    rr:propertyObjectMap [ rr:property a:title ; rr:column "Title" ; ] ;
    rr:propertyObjectMap [ rr:property a:year ; rr:column "Year" ; ] ;

The only role of the mapping to a logical table is to generate a URI from the ISBN; all the other cells are, conceptually, simply copied on the logical table. The rest is fairly straightforward.

The missing trick is to combine, i.e., to “join”, the two tables on the graph. R2RML has a separate construction for that, referred to as “mapping” the foreign keys. The following additional statements should be added to :Table1:

    rr:foreignKeyMap [ 
       rr:key a:author ; 
       rr:parentTriplesMap :Table2 ; rr:joinCondition "{child}.Author = {parent}.pid"
    ] .

Which combines the nodes defined by :Table1 with those of :Table2. And voilà! We’re done: the R2RML document is ready, i.e., an R2RML engine would generate my example table into my example graph.

Of course, there are more complicated possibilities. Triples, or whole rows, can be explicitly stored in a specific named graph, for example. Or a column defining a predicate could, actually, use a cell in another column as an object. Etc. And, to be honest, I am not even 100% sure that above is correct, I may have misunderstood some details. But the “melody” is still clear.

Note the role the SQL based mapping of the original table to the logical table has. For SQL experts, most of the work can be done there, i.e., the resulting RDF graph can be ready for further usage by an application, to be linked into the LOD, to be used with the right attributes, namespaces, etc. Which is very powerful indeed, provided… the user has the necessary SQL expertise. And, while that is obviously true for database managers, it is not necessarily true for RDF experts. For those, a slightly different model seems to be more appropriate: they would prefer to get an RDF graph ASAP, so to say, without any fancy transformation, and would then use RIF, SWSRL, SPARQL’s CONSTRUCT, etc., to turn it into the RDF graph they eventually want to have. In other words, they may not need the concept of a logical table. That is what is referred to by the group as the “default” mapping. I.e., what graph does one get if nothing is specified? If that is properly defined then, say, RIF experts can use their expertise instead of SQL. This default mapping is not yet fully specified by the group, but it is on its way; it will be published shortly, and will complete the R2RML picture. So watch that space…

October 27, 2010

Publication of the Last Call for RDFa Core 1.1

The W3C RDFa Working Group has just published the “Last Call Working Draft” for RDFa Core 1.1. As Manu Sporny, the co-chair of the group, said in his tweet, this W3C jargon is equivalent to a “feature freeze”. Ie, the group does not know of any outstanding technical issues and of missing features that it would reasonably plan to add. Put it another way, this is last round of commenting before proceeding to final implementation testing and, hopefully, to a final W3C Standard. I.e., Last Call doesn’t mean that the group takes no more comments; on the contrary, technical comments are very welcome and necessary to make it sure that the final outcome is correct. Please, send your comments to the groups mailing list: public-rdfa-wg@w3.org (there is also a public archive).

Although lots of things have been discussed in the past few months (i.e., since the last draft published in August) not many things have significantly changed, in fact. Most of the changes are editorial, making the text clearer, more precise, etc. (You can look at the “diff” file, if you are interested.) This document is for the Core, i.e., the generic RDFa processing that can be used for any DOM. It is to be expected to have, in a few days, a similar document published for XHTML+RDFa 1.1 by the same Working Group, and an HTML5+RDFa 1.1 by the HTML Working Group.

I have also worked, in parallel to the specification work, on a modified version of the RDFa distiller. While the “official” service remains unchanged and relies on the current RDFa Recommendation, there is now a “Shadow” version, that relies on RDFa 1.1. The underlying code has undergone some cleanups beyond the adaptation to RDFa 1.1 so I am sure there are bugs…

Finally, a blatant self-promotion: Stéphane Corlosquet, Lin Clark and I will give a tutorial at the upcoming ISWC conference in Shanghai on RDFa and Drupal. The RDFa part relies on 1.1… (There are links to the slides on the page but you do not expect us not to touch them any more before the tutorial itself, do you? So make sure you look at them again after the event…)

Next Page »

Create a free website or blog at WordPress.com.


Get every new post delivered to your Inbox.

Join 3,618 other followers