Inside PDF

Author’s note: Link to a Romanian translation of this article  (by Alexander Ovsov)

Data versus Information

Data versus Information

As part of an Open Government initiative the US Government has created guidelines and goals for making more of the information held by government agencies more accessible to the public. See for example, President Obama’s December 2009 directive.

Of course, I advocate the judicious use of PDF for disseminating that government information. Well, in most situations.

I will use the terms data and information to distinguish at least two ways government agencies are being asked to provide information to the public in electronic form. Data is just the raw numbers, names, places, etc. The stuff you might pull out of a database.  But data can be shaped into information by providing interpretations, drawing conclusions, pointing out inconsistencies and then packaging it in an attractive form. And then there is the most extreme Mcluhan view that the medium is the message! 

When I go to a government website I usually want information not data. When an analyst goes to a government website she might just want the raw data so that it can be interpreted, shaped and analyzed, and turned into a specific document as information not provided by the government website.

Hammers

There is the saying that if you have a hammer, then everything looks like a nail. I know for sure my 3 year old grandson thinks so when he has a real hammer in his hand.  And I have to confess that, from my view, almost all needs are best addressed using PDF. PDF is my metaphorical hammer. I have some justification for this position as the Project Leader for the ISO 32000 PDF standard.

For real, in most cases, government agencies can, and do, use PDF for effective information distribution. The information as a PDF can be downloaded, read in its electronic form, saved for later reference, shared and printed. Nearly everyone has a free PDF reader. PDF documents can also be infused with, what we at Adobe like to call, "rich document" features. The final representation of the information can be very important. As noted earlier, the medium is the message, or it certainly can make an important contribution.

However, for the person who wants raw data, PDF isn’t the right choice. See, I am willing to refrain from hitting everything in sight with my PDF hammer!  But raw data isn’t of much use unless you have metadata (data about data) describing, how to access the date elements, the ranges and meanings for the individual data items, where the data comes from, how old the data is, how authentic, and so many other important properties.  So I will argue later that, in fact, PDF can play a very important role in distributing raw data, offering a clever means to accompany that data with meaningful and accurate metadata. (I am using the term metadata here in a very general way as data about data, and maybe a little differently than you are used to.)

XML

The government initiative has surfaced people who have XML hammers. I think some XML enthusiasts, but certainly not all, go overboard. I think it hurts their cause. As I have blogged earlier (XML for •••), XML is one of the most misunderstood, yet useful, technologies we have in today’s toolkit. We need to drop our hammers and consider the facts.

Using XML for raw data is the kernel of a good idea. But here are major problems that arise because we do not go further into a more complete discussion:

1. Those who are not familiar with XML, need to realize that XML isn’t a single markup language for a single use, but it is a method for defining and using specialized markup languages. That is why we have to say XML for business cards, XML for invoices, XML for classifying political action committees and so on. There are thousands of such XML markup languages and there will be thousands more to cover all those government datasets where XML is appropriate. Please see my previous blog for more on this topic.
2. Large raw datasets can be prohibitively large when expressed in an XML markup language. Unnecessarily large, from an information theoretic view. For example, here is an XML data file that can be found on www.data.gov at this webpage. Note that, when you download this file, it is a ZIP file whose size is 11,903,362 bytes.  When you unzip it though, you get a 220,655,917 byte XML file. In this case, the EPA personnel know that XML files can be very large and have packaged it in a ZIP for downloading to reduce the transmission time by a multiple of over 18. In other words, if it takes a minute to download the ZIP’ed version, it will take over 18 minutes to download a raw XML version. After unzipping, the XML file is identical to the original. So any advocacy for XML, should always be accompanied with a filesize discussion and considerations of using something like ZIP. To do otherwise would be irresponsible.
3. XML files need additional metadata in order to make use of the data that is found within them. If you are given three numbers (007, 56, 00010) you might not guess that they represent a birthday, which is more commonly supplied as 7/10/1956 in the US.  But even if it is provided in that form, it isn’t representative of how it is often expressed in other countries as 10-07-56. Unless given additional information there is an ambiguity between the 7 and the 10 as to which is the month and which the day. And this is a trivial example. We need extra information, e.g., metadata, in order to be able to interpret data accurately and make proper use of it. The basic syntactic rules used for a markup language can and should be provided by offering an XML Schema (.xsd file) but even that does not go far enough. We also need to explain the semantics which usually requires a technical document.
4. There are other raw data formats that might be more suited to particular needs such as standard spreadsheet files (.csv — character separated values) or the Microsoft and Open Office spreadsheet formats (.xlsx and .ods, respectively).  The latter can also include the formulae and rules needed to compute dependent and summary data. In addition there may be very application specific files, not in XML format, appropriate for specific needs. For example, shapefiles (.shp) for holding spatial feature information as in the www.data.gov webpage reference earlier.

Note:  There can be an argument made that compressing/decompressing files is so time consuming that the time lost there is not made up in reduced transmission time. With today’s lighting fast CPU’s the compression and decompression times are relatively minor, but the transmission times can be a problem if you don’t have the latest and greatest Internet connections. So nearly always, the choice to used compressed data is the right one for data expressed in an XML markup language.

PDF instead of ZIP

In a previous blog (PDF File Attachments), I described the file attachment features of standard PDF (ISO 32000). To summarize, any number of file attachments can be embedded into any PDF file. They can be extracted for use by anyone receiving the PDF file. In addition, those file attachments may be compressed using the most common compression method that is used in ZIP files: deflate/flate. For the purposes of distributing government data, this is nearly ideal. The PDF file can carry the .xml or other raw data files as compressed attachments and that base PDF document, itself, can provide all the additional semantic information that would be needed in order to make use of the data — the metadata. If the raw data is in XML form, then a compressed XML Schema file (.xsd) can also be attached to the PDF document. So when using PDF, the points made above are addressed: filesize, necessary metadata to define the XML markup language used, and formats other than XML.

I have created a sample PDF envelope starting from this government dataset.  You might want to look it over. Note that both the XML dataset and the associated Schema file are attachments to the PDF that helps to define the XML markup language used for this file. I took the general introduction from the government web page and made up a brief description for each of the XML elements found in the file. Make sure to use a PDF reader that can display the attachment annotations and that can extract the attachments. Adobe Reader can do that.

Another important benefit that just works out when using PDF envelopes, is that the digital signature technology available for PDF files also can cover the attachments automatically since they are an official part of the PDF file. Government agencies can send digitally certified PDF files containing data files and their customers can authenticate that the PDF, and all the attachments, came from that agency and have not been tampered with. See my previous blog about Authenticated PDF Documents.

Other ways to use PDF attachments for government information delivery

So we described how we can use PDF to provide a complete package for raw data downloading.  But we can also augment richly formatted PDF information files with the raw data files that were used to create the PDFs.  For example, if a document contains, pie charts, bar charts or tabular information, the raw data that was used to create that formatted information can be attached to the PDF file. An annotation can be placed on the chart or table that allows the appropriate attachment to be extracted.  This is covered a bit more in my previous blog.

A third use for PDF attachments is to create hybrid files.  These are PDF files that attach the original editable source document (e.g., .odf or .docx file) that was used to create the PDF. In a sense, this make PDF editable, something that people ask for. Quoting from the OpenOffice.org website: "A hybrid PDF/ODF file is a PDF file that contains an embedded ODF source
file. Hybrid PDF/ODF files will be opened in OpenOffice.org as an ODF
file without any layout changes. Users without this extension can open
the PDF part of the hybrid file with their PDF viewer." Adobe Acrobat’s Microsoft Office tools can also create PDF files with the Office file that created them as an attachment.

So hammer away, you PDF enthusiasts.

Jim King (jking@adobe.com)

What does it mean for a PDF file to be certified as authentic?  The short answer is that it has been digitally signed by the authority that created it. I have written several previous blogs about digital signatures. PDF supports two kinds of digital signatures: approval signatures and certification signatures. Any number of approval signatures may be applied to a PDF document but only one certifying signature may be applied and it must be the first digital signature. Approval signatures are used in the same manner as the ink on paper signatures we are all familiar with. Certification signatures are considered a part of creating the PDF file so only occur once at the beginning.

Today we are only going to discuss certification digital signatures. The idea of a certification signature is to make sure that the document is authentic and has been unaltered since it was signed by the authenticating party. Don’t you want to know that the PDF statement you receive from your bank did, indeed, originate with your bank and hasn’t been tampered with since they created it?  Of course, and that is what certification signatures are all about.

One other important thing to note is that any file attachments within the PDF document can also be covered by the same certification as the PDF file itself since those attachments are part of the PDF contents. This offers a very cool way to deliver information using file types other than PDF, by enveloping them in a PDF file that has been certified. And as explained in my blog about attachments the attachments may be compressed when attached to the PDF file making them smaller for transmission. This is especially cool for the delivery of certified XML data files. The enveloping PDF can also have pages describing the attachments and how to process them.

Example walkthrough

It is instructive to use a real example to show you about authenticated (certified) documents. The US Government Printing Office (GPO) delivers some certified PDFs from their website. This is a quote from the GPO home page:

"The U.S Government Printing Office (GPO) provides publishing &
dissemination services for the official & authentic government
publications to Congress, Federal agencies, Federal depository
libraries, & the American public."

The GPO certifies documents using PDF digital certification signatures and then puts those documents onto its website. Here is one you can look at and which I will use as an example.

If you open it in Adobe Reader either within your browser or in Adobe Reader directly you should see this:

Other PDF readers should also display some kind of verification information when a certified file is opened. If not, get a better reader.

Using Adobe Reader, the blue band at the top will display when a document has been certified with a digital signature as this one has. The blue band is normally not there. What the band displays is
“Certified by Superintendent of Documents <pki support@gpo.gov>, United States Government Printing Office, certificate issued by GeoTrust CA for Adobe.”
The blue band also displays an award-type ribbon on the left end to indicate that the document has been validated, just now when opened by Adobe Reader, and found to be authentic according to the digital signature that the GPO applied. 

Special Note:  The phrase "certificate issued by GeoTrust CS for Adobe" means that the GPO obtained a signing certificate from the certificate authority (CA) GeoTrust. The "for Adobe" has alarmed some people, but it just means that the certificate references the Adobe root certificate that comes with Adobe’s products. Adobe and GeoTrust, as well as many other CA’s, have a written contract that allows those vendors to reference the Adobe root (requires a secret key from Adobe) in exchange for guarantees on the level of integrity they enforce when issuing certificates. The bottom line for someone seeing this blue band with that reference means that the signing agency had to meet strict requirements to get that certificate and that the document authenticated successfully. This does not mean that Adobe has any lock on such certifications. It is OK for some other vendor to put the Adobe root certificate into their root store and also be able to validate these documents. Anyone can purchase such a certificate from many different vendors. This page gives a lot of information about certifying documents.

Another Note:  In July 2010 I gave an introductory talk about the public key infrastructure (PKI) and digital signatures at the American Law Librarians Conference. You may want to consult that for more details about this signing technology.

Double Checking
As shown below, using Adobe Reader, a panel will open on the left when you click on the blue band at the top right end where it says “signature panel”. Within that panel you can open the various topics to see the status of the integrity of the file and of the certifying digital signature. In this case, below, you see that the check showed that the file hasn’t been changed and the signature checks out OK to be that one really belonging to the GPO because it cryptographically chains up to the Adobe root.

The workflow

The workflow that I have referred to that GPO uses with these PDF files is one where they put the certified PDF file on their website and the user opens it in a web browser or downloads it onto their disk for viewing. So the validation of the certification happens on the user’s machine using Adobe Reader, Acrobat Standard, Acrobat Professional or any vendor’s reader as the document is opened. Some readers may not do this checking automatically or at all. If you want to have the document authenticity checked, make sure your reader does that, get a reader that does or use the Adobe Reader.

Here are the details of how those PDF documents are read and check for authenticity instantaneously as you open them on your machine.
A customer of the GPO downloads such a certified PDF file. She opens it in Adobe Reader either within a browser or not. The steps that the Adobe Reader takes before showing the blue banner are: it computes the cryptographic hash of the PDF file as downloaded onto her disk. The encrypted digital signature inside the PDF file is decrypted using the signer’s public key (which is also provided inside the PDF file). The decryption can only be successfully accomplished if it had been encrypted by the private key corresponding to the signers public key (the one belonging to the GPO). The cryptographic hash that was put into the digital signature by the GPO is compared with the one we just computed. If they match then the document hasn’t been changed since the GPO signed it. If they don’t match then some tampering has happened. This is reported to the person opening the file in the blue band at the top.

Further, the certificate for the signer is checked to see if it “chains” successfully up to the Adobe root (or some other root that has been agreed to or that is commonly available). This chaining operation also uses the public/private key pairs and encryption to verify that each certificate was indeed the one issued by the CA that it says it was issued by. If all this works, then we know that whomever is identified as the signer that certified this document is indeed who they say they are according to the CA’s. The success or failure of this is also reported to the person opening the PDF file. In some cases the software opening the file may attempt to go onto the Internet to check with CA’s to make sure that the certificates are still valid and have not been revoked.

In this example, the GPO obtained a certificate from the GeoTrust Certificate Authority (CA) which chains to the Adobe trusted root.

I took that file and modified one byte in it. Now when I open it I get this in the blue banner indicating that it has been tampered with (is invalid).

Making certified documents

Here is how those PDF documents are made and put onto the GPO website:
At the GPO, the files are signed with a certificate owned by the GPO and issued by a Certificate Authority (CA) who issues them against the Adobe root. If you have Adobe Acrobat Professional and a certificate you, too, can perform this certification step. Just go to the menu item Advanced->Sign & Certify and pick either Certify with visual signature or Certify without visual signature. This example was certified without a visual signature. You need to have obtained a digital certificate that contains your public/private keys and identifying information in order to do this. These are obtained from CA’s like GeoTrust as shown here. Adobe, through its LiveCycle server product also provides services that can certify documents as they are created on a server.

The Adobe software, whether Acrobat Professional or LiveCycle server, saves the file to disk, then computes the cryptographic hash over the bytes found on the disk with the exception of the bytes where the digital signature will eventually be placed. It then encrypts that hash using the certifiers private key (GPO private key) and incorporates the result into a “digital signature” block as defined by various standards organizations. It then inserts this block into the hole that was left in the PDF file on disk to receive it. The digital signature contains the signer’s certificate that includes her public key and the identity of the CA that issued the certificate. Any other certificates that will be needed to confirm the validity of the signer’s certificate are also included in the digital signature or in other places in the PDF file.

This signed/certified document is then placed onto the agencies website (e.g., GPO).

Here is a version of the GPO story.

Jim King (jking@adobe.com)

Attaching files to PDF documents

One of the great features of PDF is its ability to carry attached files, just as e-mail messages can carry attached files. Any kind of file, and any number of files, can be sucked into a PDF file. These are held internal to PDF as “stream” objects, one of the basic 8 object types from which all PDF content is built (numbers, arrays, strings, true, false, names, dictionaries and streams). Streams start with a dictionary object but then carry along an arbitrarily long sequence of arbitrary 8-bit bytes. Stream objects meet the generic description for disk files quite well.

Various compression/decompression methods can be used on streams, so even though including attachments in PDFs can tend to make the PDF files large, it is mitigated considerably by using one of PDF’s 8 compression methods. Flate compression, the technology also used in ZIP and PNG, is most commonly used for PDF attachments since it is lossless (you get exactly the same bytes you started with from a compression/decompression cycle), it gives good compression on arbitrary byte sequences and compression and decompression are both fast. Examples which I have, go from 1/3 reduction in size to 1/200 reduction.

So what might PDF attachments be used for? One use is for “hybrid files”. The office products from ODF while making a PDF file can attach the original source ODF document to it. They have made their software recognize these special PDF files, which they call hybrid files, and when asked to open them, they actually open the attachment. In this way a hybrid file is both a final form PDF file and an editable office document. Adobe’s Acrobat products can attach Microsoft Office documents to the PDF files produced by them automatically, also. The user has to extract the Office document from the PDF, themselves, in order to then edit it with the Office product.

Another very cool way to use attachments is as if the PDF was a file folder. In fact, the ISO standard defines a way to include an index for those file folders so that when opened, a panel alongside the base PDF display can show a list, much like an e-mail list, of all the attached files, providing their name, data of creation, author, or whatever else the creator of the PDF wanted to include in the index (see 1.a and 2.a below). These are called PDF collections.

Choices for using attachments

There are basically two ways, each with two ways, to manage attached files in PDF.

1. Attachments to a base document.

1.a. The names and sizes of all documents attached can be pulled out of the PDF file and display in a special viewer panel. From this panel attachments can be extracted, additional files may be attached, or existing ones deleted.

Here is a screen shot of a sample of what you will find in the base attachments panel of Adobe Reader and Adobe Acrobat. The contents and display of this panel is fixed by the application and just includes the basic information normally available about a file.

1.b. In addition to being an attachment as in (1.a), an attached file can be represented by an annotation on a page. Various icons annotations can be used to represent the attachment like a thumb tack, chart icon, or paper clip. When the icon is double clicked upon, the attachment can be extracted or deleted. Of course, these are normal attachments that show up in the list provided in 1.a above. I like this choice because, for example, you can attach the raw data that was used to produce a pie-chart and have a clickable annotation right alongside the chart to extract the data. You can do that for all 20 pie-charts in the document. In that way, if someone wants to play with the numbers in a spreadsheet or make their own chart of the different type, they can. This is a screenshot of what that can look like with an annotation popup window showing:

            

2. Attachments with additional indexing and navigation information

2.a.  As noted earlier, the ISO PDF standard provides for the author of the PDF file to include a table of terms and values in order to be able to display a specialized index of all the attachments. PDF files set up this way are called PDF Collections. The mortgage industry has found this to be a very useful way to bring together, into one PDF file, all the various documents that represent a complete mortgage deal.  And they can be indexed, dated and named to help in getting to the right sub-document efficiently. Very similar to the base panel of (1.a), these allow the document author to use any index terms and values. In this example a “Birthday” field has been added to the basic indexing data.

2.b.  Based upon Collections (2.a), an author can also include a graphically rich interactive “navigator” that allows the creative juices to flow for showing various aspects of the attached Collection files. These are called Portfolios. This is a new feature that makes use of Flash embedded within the PDF file and called a Navigator. It has access to the indexing information described in 2.a and it can also get thumbnails of the first pages, etc. In Adobe Acrobat there are a set of pre-made Navigator files that present various effects like a carousel of sub-files, a linear fly-by of them, etc. Adobe currently supports this in Acrobat 9.0 and greater via an Adobe extension to ISO 32000-1.

PDF attachments can be a very effective way to have one file that contains a wealth of related content.

Jim King (jking@adobe.com)