Using the Chrome Debugger Tools, part 7 - the Resources Tab

The evolution of the world wide web from relatively simple web "pages" to arbitrarily complex web "applications" has been an interesting one. Whereas the early web looked more like a never-ending magazine or newspaper, the persistent trend in web design has been pushing more and more toward web pages which behave like traditional desktop applications. On the one hand, a simple article-style web page (like this blog, for instance), is server-dominant — the content is written, reviewed and published to the site's server but is, for the most part, "frozen" after that point. Every requester sees the same thing. In contrast, web "applications" are more personalized: what each user sees is based on that user's past behavior on the page. However, one of the major design points of the web — and part of its appeal to early adopters — was its "statelessness": each request was independent of any request that came before it. Any form of personalization — even something as mundane as displaying a user's name at the top of each page — needed a way to associate each request with all of those that came before it, from the perspective of a single user. As soon as users and site administrators began to push for more personalization, the need for some form of client-side storage became evident.

Some early web page authors figured out a sneaky way to "trick" web browsers and servers into pretending to be stateful even though the HTTP protocol didn't actually allow for it — they added a unique token to each URL. When the server received a request for a page with such a token, all of the links inside that page were rewritten to include that same token, ensuring that each request could be correlated to the one that came before it, from the perspective of a single user visiting the site one time. This way, the server could play personalization tricks to provide the user with a seamless illusion that the entire visit was part of one long session.

Viewing cookies with the resources tab

Tokens and URL rewriting had their limits, though. For one thing, if the user closed the browser, the token was lost. On the flip side, if the user bookmarked a page with a token in its URL, the state-tracking mechanism built into the server could end up being confused when a session from possibly six months ago appeared to have been resumed. The first "full" version of HTTP — HTTP/1.0 — provided site administrators and web page designers with a slightly more robust, but still blunt, tool to correlate clients to page visits: the cookie. In theory, a cookie worked like URL rewriting, but rather than being appended to links, it was tracked in an HTTP header. This way, the browser could automatically manage session tracking on behalf of the user so that the site administrator didn't need to ensure that every link was properly rewritten. Even better, metadata could be associated with each cookie, so the site administrator could declare, for example, that one cookie should disappear when the browser shut down, but another should persist across reboots of the computer.

The origin of the term cookie to mean a piece of data that identifies a particular user is unclear. Folklore has it that it evolved from a comparison to fortune cookies that keep "important" information inside something else.

The Chrome debugger Resources tab helps web app designers keep track of their application's use of client (browser)-side storage. Although HTML5 provides more modern alternatives to cookies for long-term storage, cookie management still remains an important part of most web development. If you open Example 1, it will set 5 cookies on your browser (if you have cookies enabled). Don't worry, I won't share any information with any third parties!

Figure 1: Cookies response headers

As illustrated in figure 1, above, cookies are set by a page through HTTP response headers. Here I've opened up the Network tab to see the "raw" data. When a web site wants to set a cookie on a browser, it transmits a Set-Cookie header with the response. The client is then supposed to remember the text that was associated with that header and transmit it back whenever (by default) any page in that domain is requested. At a minimum, each cookie has four attributes associated with it: its name, its value, the domain to which it applies, and its expiration date. The name and value must be explicitly provided (they're the first two tokens after Set-Cookie) and are optionally provided by all other attributes. If not explicitly provided, the domain will be defaulted to that of the sender, and the expiration date will be defaulted to last until the browser is shut down. The domain value of a cookie tells the browser when it should be sent back to the server. So, for the cookies in figure 1, since the domain was not specified by the response header, they will default to commandlinefanatic.com (i.e. the originating domain), and will be sent back with each request to that domain. Conversely, the expiration date tells the browser how long to keep the cookie; after the expiration date, the browser should delete the cookie and stop sending it to the domain.

Figure 2: Cookies view on response

Figure 2 shows how cookies are summarized in the network view. In spite of the many advances that HTML 5 has made in client-side storage, cookies are still an important part of modern web architecture, and Chrome continues to feature them prominently in its web debugger tools.

Figure 3: Resources cookie view

If you turn to the resources tab (the subject of this post, after all), you see a recap of what's shown in the network tab (figure 3). One benefit to the resources view when looking up cookies is that, if you load a page without the debugger tools open and bring it up it later, the resource view will show you the cookies that apply to this page — the network view will not show you this unless it was running when the page is requested. Notice also that in figure 2, the Domain tab is empty, which indicates that the server didn't narrow the domain. However, in figure 3, the (defaulted) domain for each cookie is displayed even though it's implied in the network view — the difference is that the resources view is designed to give out information about persistent storage whereas the network view is geared toward showing you what happened just now. Notice that the resources tab does allow you to remove cookies, but doesn't allow you to edit them. If you right-click on the cookies view, you can delete cookies individually, or remove all the cookies for the given domain.

In essence, Chrome has just associated those five cookies with the domain commandlinefanatic.com — every time you make a request to any page in that domain, those cookies will be transmitted as request headers, shown in figure 4. It's generally up to the server side to interpret those cookie values meaningfully, although they are accessible to Javascript programs as well.

Figure 4: Cookies in request headers

Viewing local and session storage in the resources tab

Although cookies were the dominant means of associating state and data with web clients for decades and continue to find use in web applications, they themselves weren't without their own flaws. For one thing, cookies are associated with domain names and transmitted by browsers whenever a request was made to the corresponding domain. It didn't take long before advertisers jumped on this coincidence and began taking advantage of it to track users across multiple web sites. Although this was purportedly done to provide "better" targeted advertising, it made a lot of users understandably nervous to have their browsing history centralized this way. They began to demand — and received — the ability to disable their cookies, so site administrators had to implement cookies alongside URL rewriting in order to provide page personalization. With the advent of HTML5, the web has begun to offer page designers a lot more flexibility in how they can safely and robustly (while still honoring their user's privacy) store data on the client side. Mechanisms like Session Storage and Local Storage are nearly ubiquitous now, and more and more page designers are beginning to take advantage of them.

Session and local storage are similar in terms of how they're presented to the page author — the primary difference is that session storage disappears after the user closes the browser, whereas local storage persists across multiple restarts. Otherwise, both types of storage are simple key/value stores — you can insert new key/value pairs and retrieve the values associated with keys later. Example 2 is a simple use of local and session storage. If you open it and input the data shown in figure 5, you'll see those inputs reflected in the session and local storage areas in the Chrome debugger tools as shown in figures 6 and 7.

Figure 5: Storing some values in local and session storage

Figure 6: Viewing local storage in the Resources tab

Figure 7: Viewing session storage in the Resources tab

Now, if you shut your browser down, restart it, and navigate back to example 2, you'll see that local storage still shows the stored value, but session storage is blank as expected. However, if you simply navigate away and then navigate back without restarting your browser, the session storage values are retained — session values persist for the lifetime of the browser session. On the other hand, if you open an entirely new tab (or window) and navigate to the same page without shutting down the browser, you will see a new, completely independent set of session storage variables — sessionStorage is "tab aware" whereas cookies are not. Of course, one potential downside of using local or session storage is that it isn't, by default, accessible to the server, so you may be forced to move certain functionality into the client.

From the local storage resources tab view, you can add or edit local storage values; double-click an empty line to add a new one, double-click an existing line to edit either its key or its value. The (x) at the bottom of the screen allows you to delete a storage item — be careful, since there's no undo function! This is particularly important for page authors to be aware of — since the user can (easily) change or remove any local storage values, they should never be relied on for secure functioning (nor, of course, should cookies, which can also be modified, albeit with a bit more effort).

Note that session storage is meant to be segregated between multiple windows or tabs whereas local storage is stored long term and also shared among different tabs. This does introduce the possibility of conflict if two tabs are updating the same local storage entry concurrently. Therefore, the local storage standard allows pages to register to be notified of storage events that occur in other windowsexample 3 updates example 2 to keep the local storage values in sync (note that there is no way, nor should you want or try to, keep session storage values in sync). If you open example 3 twice (in two separate tabs), any change you make in one will be shown in the other. Interestingly, as illustrated in figure 8, if you go to the resources tab and make a change there, that change will be reflected both on the other tab's page as well as in its own resources area.

Figure 8: Keeping local storage values synchronized between tabs

Viewing Web SQL databases in the resources tab

Session and Local Storage are more Javascript-friendly forms of more traditional HTTP cookies — they manifest themselves as key/value pairs and don't offer the app developer a means of relating one to another automatically. In contrast, for more complex applications, HTML5 created the concept of client-side databases modeled after their successful server-side counterparts. The first such effort was the proposed Web SQL standard that was exactly like a traditional relational database such as Oracle or MySQL.

In spite of the fact that the Web SQL standard has been deprecated by the W3C, Chrome still has pretty extensive support for it. I'll cover it briefly here since you might come across it, but you shouldn't write any new functionality around it. Example 4 will create a web SQL DB when you click the Open button, executing:

  var db = openDatabase('testDB', '1.0', 'test database', 1024 * 1024);
Once you do, the database appears in the Resources tab under the Web SQL drop-down as shown in figure 10.

Figure 9: WebSQL view with no databases

Figure 10: WebSQL view with one database

If you click on the newly created database, you'll be presented with a miniature console as shown in figure 10. From here, you can execute SQL transactions and see their results right away. So, for instance, type

create table keyValueTable (key varchar, val varchar);
to create a table with two columns. Now the table shows up in the list of database tables, but is empty as shown in figure 11.

Figure 11: empty table

Navigate back to the console by clicking on testDB again and type:

insert into keyValuetable (key, val) values ('abc', 'xyz');
Now if you navigate back to the keyValueTable, you'll see the newly inserted data as shown in figure 12.

Figure 12: key value tale with one row

Notice that Chrome created a rowid column to act as the primary key for your newly create table. If you had created a primary key (by marking one as primary key in the create table statement), Chrome won't create one, but you'll see it twice in the tabular view.

The tabular view doesn't offer much functionality of its own. You can refresh the data or limit the visible columns using the controls along the bottom, but you can't edit or delete data — you can only do that via the console using statements like

delete from keyValueTable where rowId = 1;
You can also run SQL select statements like the one shown in figure 13, below, and the result will appear on the mini console.

Figure 13: select statements in the webSQL console

You may have noticed that the webSQL console autocompletes for you, which can come in handy if you've forgotten your SQL syntax. Everything that you type into the mini-console will be interpreted as though it were:

transaction.executeSQL(<whatever you typed>);
Which is how web sql was intended to be performed before it was deprecated by the W3C.

If you close the browser, navigate back to example 4, and bring up the resources tab, the WebSQL line will be empty. If you click the Open button again, however, triggering the line:

var db = openDatabase('testDB', '1.0', 'test database', 1024 * 1024);
the database will reappear, including all of the data that you inserted previously. You won't see it in the resources tab until you reopen it, though.

Viewing Indexed databases in the resources tab

So I mentioned that the Web SQL standard is deprecated and shouldn't be used — so what should you use if you have storage needs that are a bit more complex than the simple key-value pairs that local storage permits? HTML5 also introduced the IndexedDB standard, which is well supported, especially in Chrome. Example 5, illustrated in figure 14, will create and populate an IndexedDB in response to button clicks.

Figure 14: IndexedDB view with no databases

Initially when the page is loaded, no database has been created; click the Create button and a new database will be created using the code:

var request = indexedDB.open("testDB");
However, you won't be able to see it initially; the resources tab isn't dynamic this way (at least not where IndexedDB's are concerned). Right click and select "Refresh DB" as shown in figure 15.

Figure 15: Refresh IndexedDB

Now you see that the database has been created along with two empty tables (names and columns courtesy of the WWW specification on IndexedDB). These two tables were created after the database was opened via:

	request.onupgradeneeded = function()	{
		var db = request.result;
		var store = db.createObjectStore("books", {keyPath: "isbn"});
		var titleIndex = store.createIndex("by_title", "title", {unique: true});
		var authorIndex = store.createIndex("by_author", "author");
	};
Note that this code is only run the first time you create the database (e.g. onupgradeneeded) to create a new table. You can see the details now as shown in figure 16.

Figure 16: IndexedDB database metadata

This DB is associated with every page in the commandlinefanatic.com domain — you can verify this by navigating to any other page on this site. The table has two indexed columns, shown in figure 17 — it's up to the database creator to optimize it by indicating which columns should be indexed.

Figure 17: Empty table with two indices

If you now click the populate button, data will be inserted into the database using:

	var tx = db.transaction("books", "readwrite");
	var store = tx.objectStore("books");

	store.put({title: "Quarry Memories", author: "Fred", isbn: 123456});
	store.put({title: "Water Buffaloes", author: "Fred", isbn: 234567});
	store.put({title: "Bedrock Nights", author: "Barney", isbn: 345678});

	tx.oncomplete = function()	{
		console.log("population successful");
	};
Note that, in order to see this, you have to click the refresh icon at the bottom of the table view, not the Refresh IndexedDB context menu item that you used in figure 15. Like the WebSQL view, you can't edit the values or insert new ones from this view. You can do simple searches and page back and forth over a large volume of data, though, since there's no equivalent of the select statement here.

Figure 18: Populated table

The indexes themselves can be viewed in order as well by clicking on them. Again, you'll have to click the refresh icon before you see this, even though it's populated.

Figure 19: Populated index

Interestingly, if you close the browser and navigate back to this page, you will see this IndexedDB, along with all of its data, in the resources tab. This is in contrast to the WebSQL handler that only shows you the database after it has been opened by one of the pages.

Application Cache

As I mentioned in the introduction, HTML 5 represents a shift in thinking about the web from pages to applications. The HTML5 Application Cache makes this shift complete by allowing a web application to declare to the browser that it should be downloaded in its entirety and served from cache permanently from that point on. In this way, a web application like a word processor or a spreadsheet can remain functional while it doesn't have a network connection (such as on an airplane). Web pages have always been cached, but the web application cache allows the page/application author to be explicit about exactly what should be cached and how.

Example 6 is a bare-bones "application" with some cached pages. The cache control is declared in the HTML tag as:

<html manifest="art039ex6.appcache">
This is a reference to a cache manifest file:
CACHE MANIFEST
# v1 - 2015-04-27
http://www.commandlinefanatic.com/art039ex6.html

NETWORK:
art039ex6_net.html

FALLBACK:
art039ex6_preference.html art039ex6_fallback.html
If you open this example in another tab and navigate to the application cache view as shown in figure 20, you'll see a summary of the manifest file in tabular form. The application cache has three categories of entries: Master, Network and Fallback. Master entries will always be retrieved from cache, even if you're online. Network resources are the opposite — they'll never be loaded from cache, and won't be available in offline mode. Fallback entries are a sort of hybrid: if you're online, the first entry will be retrieved; if not, the fallback will be presented instead.

Figure 20: Application Cache View

If you want to see a summary of all of your offline applications, you can visit chrome://appcache-internals as shown in figure 21 below.

Figure 21: Appcache internals

Frames

Last but not least is the Frames view. Although it's at the top of the resources tab, it almost feels out of place next to so many advanced, shiny new HTML 5 additions. The frames view of Example 7 illustrated in figure 21 provides you a list of all of the "standard" parts of a web page/application — the HTML, CSS, Javascript, images, web fonts, etc. that are downloaded to provide a page to display. This will always be present and have at least one entry, since you need at least an HTML page in order to show something in a browser.

Figure 21: Frames View

For the most part, the frames resource view provides a rehash of the information from the Network tab, but gives you a more "after the fact" view of how the page is structured. If the page includes any embedded iframe elements, they'll be separated out as shown in figure 22.

If you have a page with traditional frameset elements, they'll also be separated this way. Hopefully you'll never need to know this, though, since framesets have been deprecated for a very long time.

Part 8: The Audits Tab

Add a comment:

Completely off-topic or spam comments will be removed at the discretion of the moderator.

You may preserve formatting (e.g. a code sample) by indenting with four spaces preceding the formatted line(s)

Name: Name is required
Email (will not be displayed publicly):
Comment:
Comment is required
My Book

I'm the author of the book "Implementing SSL/TLS Using Cryptography and PKI". Like the title says, this is a from-the-ground-up examination of the SSL protocol that provides security, integrity and privacy to most application-level internet protocols, most notably HTTP. I include the source code to a complete working SSL implementation, including the most popular cryptographic algorithms (DES, 3DES, RC4, AES, RSA, DSA, Diffie-Hellman, HMAC, MD5, SHA-1, SHA-256, and ECC), and show how they all fit together to provide transport-layer security.

My Picture

Joshua Davies

Past Posts