An agent can run dependently or independently.

As a dependent it is loaded just like you'd load any other program that you use. The agent then subscribes (communicates) with all nodes the hacker communicates with and has identical access that the hacker has. The agent may use any programs that are loaded onto the hacker's commlink (except agent programs) which then count against response limit.

As an independent it is loaded onto a node (any node) and runs from there. It subscribes (communicates) just as any other matrix user does (hacker, node, or drone) and must hack access or be given access to all nodes it enters. It can access multiple nodes at once and has its own matrix access ID. It can use any programs that it has loaded or that it can load from the node on which it runs. Whatever node it is currently active in determines it's response.

IC are treated as either dependent agents (loaded onto the host node) that only runs if an alarm is triggerd or as independent agents (in which case there is only a semantic difference between them and agents). Once an agent is running it follows whatever its orders are like a drone (which may include subscribing and accessing several nodes). All IC traces back to the host node.

Cetiah,

The simple fact is that I believe the word "icon" means what the word "icon" means; a symbolic representation of something else, nothing more. I have not found anything in the SR4 text that gives an icon more meaning than that.

Target icons in matrix combat only refer to IC, agents and Persona because they are the only "active" objects on a system other than an OS, which can be crashed but not attacked. Passive objects (data files) can be deleted but do not require attacking them to destroy. It's not D&D, you don't need to whack the chest with the sword to break it. Use Edit and wipe the whole file.

I've made modification to the summary/conclusions above.

Thanks to kigmatzomat, I understand now that I need to EDIT the summary, not stab it with my knife. That saved me so much time.

I think it's probably more like a sliding scale.

SR4 Agent 1-3 is probably like a SR3 Smart Frame
SR4 Agent 4-6 is probably like a SR3 Agent

The difference in SR3 between the two was just capability. Agents were far more capable than a smart frame.

@Moon-Hawk:

A rating 6 agent definitely beats a hacker 3. But to me the way I presented it makes the rules for agents, hackers, drones, and IC the same so it is easier to use for me.

So if a host is set to auto-restart crashed IC, does that mean that a hacker could never defeat it?

Or would crashed-then-restarted IC (and thus agents) restart in a "default" way? If yes, then I suspect the average IC is not defaulting to "Blitzkrieg" at all times and, thus, a crashed-then-restarted IC would go back to analyzing or other less active, for lack of a better term, actions.

I can see that, but then how do you resolve the Schroedinger's Agent problem? I think we agree that you shouldn't be able to restart Schroedinger's Agent (regardless of our reasoning) but if Schroedinger's Agent were to be crashed, does it respawn on its "owner's" commlink? If so, how does it know where the commlink is? How does the commlink know what happened?

What about the middle ground (and this also solves Schroedinger's Agent): agents crashed while operating as a part of a persona are not lost but independent agents are. A hacker can track an agent's datatrail to the node in which it was crashed to recover it, but if that datatrail is lost, deleted, spoofed, redirected, etc. and said hacker cannot find the resting place of said agent, then the agent is irrevocably lost. Clearly, if an independent agent is acting within the same node as the persona, then no tracking is necessary.

Howabout this: When "Schroedinger's Agent" is on a node, and that node is disconnected, the agent carries out whatever it's instructions were, and then waits around for a reconnect. The hacker who loaded the Agent gets a choice, when the agent goes offline, to either keep that subscription link "open" and unused, or to use it for something else. If he uses it for something else, then the Agent loses it's connection with him and becomes unrecoverable. If matrix access is later reopened, the agent may begin roaming about the matrix, as a "free agent"

I like tangents as much as the next guy, as long as the original issue has been addressed and it stays on an SR topic. But let's not waste space on arguing about arguing or little post shots, 'k?

Sometimes just arguing your point helps you define your own thoughts better, so that makes sense. And besides,even if you don't agree with others, different view points are always a good thing. They give you more to think about and help refind a point of view. Me, I'm just glad someone's happy after 11 pages.

According to the FAQs definition of Subscription versus Access I can jive with still being able to keep your subscription list, including an independent agent, when you get crashed.

A good analogy would be when subscribed to 3 nodes, an independent agent, and a drone. What happens to your subscription to the drone when you are dumped from the matrix?

According to the RAW in the FAQ subscription means communication. So you still communicate with the drone even after being dumped because you still have a signal/connection with the drone. Crashing doesn't knock it off your list.

I guess the question becomes: what exactly happens to you when you are crashed on the matrix?

That sums up the big difference. Your base assumption rejects the current model of computer operation as being relevant, which eliminates a large number of common-usage defintions for common english words. Since the SR4 text lacks explicit definition for these terms, you are forced to infer definitions and do so in intentionaly disregard existing context.

IMO, that is comparible to rejecting the current lexicon of firearms and having to infer a new definition for "fire," "magazine," or "revolver." Both are equally integral to the game.

If this were a pure fantasy game (D&D, Earthdawn, RuneQuest) or a significantly alternate quasi-modern world (DragonStar) I'd accept a rejection of real-world models. However since SR4 is explicitly stated to be based on the "real world," at least until the 1990s, it is patently unwise to reject all current models of technology.

I would dispute that I miscategorized your position with the house of cards. You refute, in essence, any similarity to real-world weather conditions and those that exist within the game and, since it can be inferred that the big bad wolf is dead, assume that houses of cards will not be blown down.

Okay, dipping my toe in the water for the first time here.

pg212, core rules

To a certain extent, kigmatzomat, isn't that what we do with firearms, though? The RAW works by abstracting the real world concepts into things that might exist 70 years from now. In keeping with our meteorological examples, I notice that in the real world, wind is a major problem for shooters when firing at long distance, but that the RAW have decided that this modification is not necessary to represent within our abstraction.

I completely ignore all real-world concepts with respect to any game understanding that the game's abstraction may redefine those terms. However, when the game's abstraction does not redefine those terms, it is only at that time that the real-world's context begins to apply.

However, Dashifen, range does affect accuracy, and it could be argued that a smaller target profile, combined with windage and elevation adjustments, are factored into the blanket dice pool penalty for range, could it not?

*EDIT*
Because I can't fricken' spell.

Cheops, I would respectfully disagree with you there. Your commlink can have a subscription list of 1000 nodes. I think it's a peice of terminology that should be kept thematically separate from subscribing. I would argue that in context, the act of subscribing is something that is actively done, while a subscription list is passive. A lot like a list of "trusted" sites that your firewall won't automatically restrict access to(access and firewall in the real world, not SR4 speak.)

As for the FAQ, in general I don't find it very helpful, as often times it seems to directly contradict the rules as they appear in the book, making it more of an errata than a FAQ. If you would like me to give some concrete examples of this, I'll try to dig em up. But for the time being I'm going to assume we agree that the FAQ is poorly named. In my personal opinion, I like the book when combined with the errata(flechette really needed that +5 to AP) but I find the FAQ one has to pick and choose.

However, even looking at the context of the FAQ text, I would say that when they say "subscribing" what they are describing is closer to the "active subscription" from the text than the "subscription list" portion. Do you agree with this?

Not to be a dick but the FAQ comes from the DEVELOPERS of the game. So any rulings or clarifications made in the FAQ, no matter how flawed or how much you dislike them, are OFFICIAL. And therefore RAW (Rules As Written).

In this case the FAQ is very clear as written.

I do like your distinction between subscription list and subscription. That makes it a little easier to understand how it works.

As for the FAQ description I don't agree with you. Here's an analogy. You show up at a party. At first you are in the doorway all by yourself. You make your way around the house saying hello to everyone. Eventually you have made your greetings and gravitate towards the room with the people you are friendliest with. You find a seat and start talking to the people around you. Over the course of the evening you change chairs, conversations, and even rooms until eventually you are back in the cloak room getting ready to leave.

Think of the entrance as logging on. Your icon then travels around to the different rooms (RTGs) talking (subscribing to various nodes). While you are at the party you are maintaining all the people on your subscription list. However, you are only accessing the people who are sitting next to you. You can talk to anybody in the room (or even the next) by shouting across the conversation at them. However, access to people at the party is easiest with those nearby. So you have access up to the number of people who are sitting next to you (determined by the size of your system).

The FAQ doesn't describe subscribing as anything more than just saying hello to each node. Subscribing is just a handshake.

With the description in the FAQ I'd take the line in the book "actively subscribe to (access)" to be a clumsier way of saying "you may only have access to system x2 nodes, agents, or drones at a time." Active subscription = access. Subscription does not = active subscription. Active subscription are the people you are actively engaged in conversation with. Subscription is just the people at the party. The FAQ gives the distinction.

It even goes so far as to define the difference between people you know and those you don't. That's what access is. You talk to someone you don't know (subscribe). You introduce each other and have some things in common. You begin to actively subscribe to that person. Access is granted because you both like the same things. Now you go and talk to the hot chick in the corner. You subscribe to her as you approach but she greets you icily. Access is denied. You can leave her on your subscription list and continue to recognize the fact that she is still in the room or you can completely ignore her and take her off you list.

Okay. So revised thoughts:

1) Subscription list- This is a list of nodes that you have encountered and decided it's okay for you commlink to attempt to access. Whether or not you actually are allowed to access it is a different story, but you have a list of nodes you know about. I imagine a similar parallel would be a list of IPs.

2)Actively subscribing- This is what I don't understand. I believe that when the developers say "actively subscribing" they mean exactly the same thing as " subscribing." Unless maybe they meant access, in which case they managed to muddle the waters mightily.

3) Subscription- The first step of accessing a node. As the FAQ states, it is the act of "creating and maintaining a connection between two nodes." Essentially, subscribing to a password protected node would get you to a login screen.

It seems to me that subscribing, actively or otherwise, can be done irregardless of the number of nodes your signal(whatever you want to call it, icon, persona, etc) has to hop through to reach the node you are subscribing to. From the text as I read it, especially considering the FAQ that says you don't need to hack every single node in a daisy chain, when you subscribe to a node, you're ignoring all the nodes in between you and it. Subscribing limits apply only in parallel, when you're trying to access multiple nodes at once(keeping four tabs open on firefox), instead of moving data(somehow) through a series of nodes to say, place a comm call.

The only way I can make all three variations of subscription work, in my mind, is if I say, "Subscription is the initial handshake process." and "Actively subscribing is accessing a node after having subscribed to it." However, if this is the case, you can only ever access something after you subscribe to it, so there is no difference between active subscription and access, except semantics.

And finally, the thing that really messes me up, is the part in the FAQ that says you
don't have to hack through every node in a daisy chain. In my mind, this somehow means that if I have some way of knowing the address, or access ID, or whatever it is that lets a node be on my subscription list, I don't have to hack my way past any ice on a corporate server except for the stuff on the node that has my paydata. Now, that CAN'T be right, yet it says that

FAQ, Hacking section

@kigmatzomat:

I believe you are correct that access = active subscription and the only difference is semantics.

I like the fact that you only ever have to hack one node now. The NodeQuest got fairly tedious and turned lots of people off the Matrix. The way I run security now is linking matrix to physical and focusing more on detection of active runs than on prevention. If security can detect the hacker they can track to his source and deal with it that way. Black IC is now fairly common in my games. 4000 nuyen gets you Blackout 4 and a -2 threshold to your track attempt.

@cetiah:

Another good measure are agents that monitor passcode lists and cross-reference with working hours lists. If any new passcodes are added or if it detects a log-on by a user that doesn't normally log in at that time it alerts a spider who can take appropriate action like tracking.

To Extend my example:

The House is the Host's persona. Firewall is the door, response are the corridors, signal is the size of the house, and system is the size of each room. The host moves around the party entertaining his guests and is thus more "mobile" than most of his guests who stay in relatively the same room (RTG) all night (like nodes).

Over the course of the night he finds that his child (agent) has snuck down to see what's going on. The host has two choices: send the child back to bed (don't use) or let the child stay up (load the agent). In this case the host lets the child stay up but tells the kid to stay near him (loads the agent to his persona). The child gains a subscription list of all the guests and actively subscribes = accesses all the guests that its father also accesses.

Later in the evening the host orders the kid to go around and refill all the bowls of chips (makes the agent independent). NOTE THAT THE KID IS STILL IN THE HOUSE (COMMLINK). However it is now operating independently although still loaded onto the hacker/host's commlink. The kid now goes from room to room refilling chip bowls.

However, whenever the kid goes into a room (RTG) to access the chip bowl (targeted icon) he has to converse with the people around the chip bowl (access different nodes to get at the target icon). The kid is now engaged in a conversation with multiple guests while he refills the chip bowl. How fast the conversation is (response) depends on how fast the guests are asking it questions (response of the accessed nodes).

EDIT: If the agent were not allowed to operate on multiple nodes as the other side of this debate asserts then the child would only talk to one guest at a time -- ignoring the others until he was finished with the current one. I don't think that's the way it was meant to work.

Eventually the child refills all the chip bowls and returns to the host (hacker). The host now decides that it is way past bedtime so he takes the child to bed (unloads the agent). The host now returns to the party.

Okay, back to the subscription thing. If Actively subscribing=access, and subscribing = initial handshake process, then how can one be subscribed in any way to an agent when one isn't online? I'm okay with being able to do stuff with your commlink when offline, but nothing that we can't do currently with no connections to outside computers. I mean, you wanna dictate a document, check you address book, or listen to some music you downloaded, no problem. But maintaining a connection to something else, even if it's with no nodes between you and the connection, both of you just using your own wireless range, you have to be online to do that. My laptop has a little switch on it. I slide the switch one way, it disconnects my wireless antenna and I can't access anything. In my mind, that is "offline."

I'm also okay with being able to pick connections back up when you hop back online. But I don't see how it's possible, or where it's indicated in the text, that you can maintain a subscription(not a subscription list) while offline.

When you are completely offline, no you probably can't maintain your subscriptions anymore. Basically you're logged off of everything (in the deepest darkest jungle or behind 3m of jamming paint).

However, the book does make a distinction when it comes to logging off and jacking out. You can jack out and still be connected to everything. Basically your persona is still there but it is motionless and does nothing. Everything still works but may start to time out or disconnect. Logging off is what actually signs you out of a node. To get completely off-line you'd have to take a simple action to log off of each accessed node. Presumably you also have to "turn off" you subscription list too but it doesn't give rules for this. I'd say probably a single simple action would sever the connection to all of you subscription list of inactive nodes.

p. 216

"Every time you are online...your presence is logged. Every device, terminal, and wired jackpoint has a unique serial number assigned by the manufacturer.... This access ID is associated with all your online transactions and typically logged by any device you access. This record is called your datatrail..."

It says right there that everything you access keeps a log of your aID when you access it (presumably not when you inactively subscribe).

p. 225 Hackers and Editing

"Note that many hackers use their Edit program to eliminate any records of their tampering or illicit activity on a node."

So the datatrail also includes all activities taken on a system.

So every month make a list of who was on at what time matched to their aID. Corporate HR has a list of all employees with associated aID's (could even have an office and a home aID for each employee so they can telecommute). Match the two and voila! A list showing which employee is accessing at what time from what terminal. If Joe Blow has worked there for 6 months and never logged on at 2 am then it is a safe bet that it is worthwhile to have the spider follow up and ask him about that.

The agent basically sits there watching the Edit logs and cross checking with the aID list. Red flags are reported to the spider. This should be a relatively easy thing for an agent to do since I imagine most reconciliation work is probably done by agent with a human accountant verifying the final result.

I would say that circumventing all the various clever tricks like having an agent monitoring the access logs are covered with the simple hacking roll and your stealth program. If we take the view that program rating are not constant, and that software is improving over time, then we have to admit that a rating 6 (or even 7, hot hot hot) programs is cutting edge stuff. They already thought of this, and it's already included in what the program does, be it a spoof, stealth, analyze, etc.

If you keep coming up with little tricks to make things more difficult, Shadowrun turns into a game of, "Read the GM's mind." I had a GM like that, and while we it made us paranoid as shit, it wasn't really any fun. It would just piss us off when we got caught again for the new thing that the GM thought up. There's so many ways to screw your players, I figure most of those are incorporated into dice rolls. I figure players should be rewarded for thinking sneaky, by basically assign bonuses to dice pools or reducing thresholds. Making the players think to check for the exact thing that you thought up to trip them up either turns into an exercise of railroading if you want them to find it, or frustration if they don't.

The worst part of this is that it slows down the game. The GM says something innocuous, and the players go to red alert cause they think it's another sneaky ploy, and it takes you 20 minutes to explain that when you described the street as strewn with rubble, it didn't mean that an ambush was imminent and they can continue on to the damn meet.