HTTP-Stateless

• Could connect to any web server in a data center, racks of machines
• TCP/IP covered next week: it is a way to transfer raw data
• What does an HTTP command start with? method (eg. GET POST etc)
• What's the header we used in HW? Host

• URLs start with a? schema
• How to know what machine to connect to? host/server name

• Sends back HTML
• How does the browser organize the HTML? What kind of datastructure? DOM/tree
• Know how to write the DOM tree
• At a high level, how would we use Javascript to change the color of the paragraph text?
• Where does the Javascript run to do this?
• At a high level, how would we implement infinite scroll on this page?
• What protocol would be used to fetch more content?

• State is information associated with some item that determines how you can interact with it
• Light switch has state: which position is it in
• Coin flips do not have state: doesn't matter what previous flips were
• We'll be talking about state at many levels: if a resource has state, if a protocol has state, building stateless protocols on stateful protocols

• Calling a bank is generally stateless. Each time you call, must identify yourself.
• Often, even when you transfer to another dept for another question, but identify again
• If you get disconnected, can you just call right back and expect to continue?
• vs. family: can continue a conversation. Generally stateful
• What is one of the main reasons for the difference? Why is one stateful,

another not?

• who specifically you are talking to. Calling a bank could talk to any individual. Very similar

• HTTP/1.1 keeps an open connection, but doesn't change the commands, headers, content required for the request
• When making requests for images on a page, still must ask for absolute path
• *Cookies will be covered in 2 weeks

• src has a relative path
• Browser appends relative path to current URL, then makes HTTP request
• HTTP request is for the absolute path

• Browser translates relative paths in HTML to absolute paths in HTTP request

• Future lecture will focus on managing state in DBs
• In general, you need multiple disks to dependably store state
• Stateless means you don't have to coordinate
  • Easily scale without coordination
• Authentication: must send credentials each time, but browser mostly handles it

• Associate with an IP? No, what about users from the same building?
• Where should we send authentication information in a request? (for normal pages, not login form
• Headers: metadata about request, eg. who is sending it?

• Databases are typically used to store info, we will cover them later in class
• When looking at an individual request, we send all required information
• Just like we build styling on top of semantics (CSS on HTML), build stateful application on stateless protocol
• Like a telephone call (pre-mobile): you have to introduce yourself each time, but after introducing yourself I can remember details about you

• (Transition)
• We talked about response codes last week, here are a couple fun ones

• Unfortunately, a few things can mean "dynamic"
• web pages (what we're talking about, generated by server)
• DHTML: basically javascript (an attempt at a buzzword. We're not talking about that right now)

• No computation needed
• RAM can "cache" frequently needed pages
• But static pages can't change (by definition)
• And static pages have to be stored with their full HTML

• Customized home page: recommendations, etc.
• New activity that you want to see. How boring would a static feed be?
• Maybe someone only recently gave you permissions to view a resource
• Getting the current weather
• Resource have state: GET request for whattimeisit.com

• Case study: different ways to implement Craigslist

• Very fast serving
• Simple to write, understand
• No database dependencies
• Easier to backup? Just copy files
• Less flexible: How do you change the template?
• Less normalized: how can you data-mine your ads?
• More space: repeat the Head HTML over and over

• Slower to serve: have to calculate response
• More complex to write: must interact with other libraries
• Database dependency: DB must be up to serve traffic
• Backup different systems: data, templates, code. How to backup DB?
• More flexible: instantly change the template
• Data separated from presentation: can mine, transform, annotate, modify data
• Less space: Just the data and the template once

• Dynamic
• Static
• Static
• Dynamic
• Dynamic

• We'll discuss caching in another lecture
• Caches are not exactly HTML files on disk
• Usually stored in RAM