I'm curious to how my view will hold up to the scrutiny of dumpshockers. I'd like to get this hammered out, as it seems the consensus of dumpshock right now is contrary to my own, so hopefully I can get some good discussion going about this.

There are 2 modes agents can operate in. SR4, 227"Agents can be loaded into your persona like other programs (taking a Complex Action), allowing the agent to accompany you to any nodes you access. Agents can also access other nodes independently if instructed to and if they either have the passcodes or are carrying an Exploit program and can hack their own way in (as independent icons)."

My interpretation of the rules is that: When loaded into your persona, the agent is present in nodes that you are present in, and is unable to go anywhere without your icon going first. The agent utilizes the response of your commlink. This method is PA in the poll.

When operating independatly, the agent is able to act like a unique persona, without any trail leading back to your commlink. The agent has whatever programs it brought with it, and must access/exploit nodes in order to use them. It uses the response of the node it is present on. This includes agent/ic spawned by a node itself, as they aren't running via any persona. This method is IA in the poll.

My reasons:
SR4, 217-218. "When you are accessing a node, you may set your Analyze program to automatically scan and detect other users/icons on that node with a Simple Action. A successful scan will be reported to you. The program will maintain that task for as long as you are on that node or until you kill that process."
This shows that accessing a node and being on a node are the same thing.
SR4, 220. "If you don’t want to maintain a connection to that node, you can run a Browse-equipped agent (p. 228) instead." This shows that agents are able to run without any connection back to your commlink. You can run a browse program on a node with an open connection that can be traced to you. Or, you can send an agent to do it, with no data trail because its operating independantly of you.

If an agent operating independatly has no connection to your commlink, then it obviously is not operating utilizing your commlinks hardware. This is why an agent operating remotely utilizes the response of the node its persona is in. Because its not connected to your commlink in any way. When you are running them off your persona, you can utilize your commlink's stats, as there is an open connection, between your persona and your commlink.

Not enough options.

My interpretation doesn't seem to be in your Poll.

PA runs off the Response of your commlink.
IA runs off the Response of whatever node it happens to be running from. This in turn may or may not be your Commlink.

So if anything, I'd like a new poll. One where I could actually answer it. As I interpret it, an independent Agent is running just like a Hacker is - they have a source node (which limits them with its response) and then they have whatever nodes they happen to be interacting with (which don't).

-Frank

I like Frank's interpretation because that means I can clean out a commlink store, copy my agent onto each of them (I have the source code since I wrote the agent program, so the copies are free). Then I can launch a horde of hundreds of agents to pwn any system I want. With no system degradation, even the best IC in the world will be wiped out in one round. Unless they know that trick and have thousands of agents running in their system...hmmm, this could get silly. When is SR5 coming out again?

My personal method of dealing with the Agent Smith problem is to abstract the actions of agents. If you have one or one hundred, you only roll & act once for any single activity (generally the one with the highest bonus), as if there were only a single agent; and having greater numbers of agents only gives the ability to multitask & it's harder to stamp them out. As an extra element, agents cannot provide teamwork benefits to anyone or anything in hacking.

The idea is that having multiple copies of a graphing calculator running doesn't make the graph any more awesome, and the rating for the program/agent should already take into account how much processing power they take up...which is a sufficiently large number within a single commlink that stacking them doesn't give an appreciable benefit (other than holding more programs before losing response).

Agreed. There are not enough options. My own interpretation is in line with Frank's. I see no sensible reason why an agent can run happily from your commlink, operating on a remote node, fighting in cybercombat, running exploits, stealth and all that jazz, but only if a human is sitting next to it holding its hand. The "independently" that Tarantula has fixed on refers, I believe, to when the agent is loaded onto a different node. That's not compulsory to be present in another node, as demonstrated by the fact that agents can do this when accompanied by your persona and by the fact that agents can be present in multiple nodes which would screw the whole theory.

It also doesn't make much sense in terms of what we know of the Matrix. Being present in a node in VR represents a connection and interaction with that node. I don't think that can be disputed. So this idea would mean Mr. Corp Exec can't have a personal assistant that goes off and books flights, arranges meetings, downloads SIMs for me, etc. unless I either he goes with it to hold its hand, or everytime his (along with everyone else's) program wants to go off and book a flight, it has to actually cease running on his commlink, upload itself onto a remote system that is willing to run it, do what it wants to do and then reverse the procedure. This makes little sense from security and efficiency points of view on both the execs and the airports side. And it also brings us back to the issue of working with multiple nodes, again.

The Agent Smith problem would not be the first loophole that has crept into the 4th edition rules. (*cough* blood spirits *cough* summoning loop *cough*), so I think the existence of a potential problem being a consequence of what the rules say is not really an argument that the rules didn't say that.

The way that I have dealt with the Agent Smith problem is part fluff and part rules. Firstly, on the corp side, I have made long arguments elsewhere as to why corps wouldn't field agent armies. On the side of the hacker, there is also the expense, but more particularly the point that stealth is almost always critically important to the hacker and it is the one thing that Agent Smith army cannot do. Often, it is important that the victim does not know that they have been compromised. There are also strong options that a target system can use when it knows it is being attacked. For a start it can sever outside connections or shut down. Either one of which, by RAW, it can do before the ASA can do anything useful; and both of which have no counter.

More so than this, the ASA is massively detectable, massively traceable. You can spend two months sneaking round the city and planting commlinks in dustbins all set to launch their attack on midnight of the 31st. You're likely to be found. You've probably gained nothing other than some mild interference in the target business (if it doesn't immediately use a different Matrix access point, that is) and in return catapulted yourself to the top of Lone Star's wanted list. Of course a big corp might not worry about Lone Star, but now we're into the realm of politics and counterstrikes and corp war.

There are other counters that a GM can pass under RAW. For example, you can treat a massive agent attack on a node as a teamwork test (which it really is) in which case the number of agents that can work together is limited by their rating. You could also argue that given the agents are all clones of each other, they're not going to be able to help each other very much:
"Hey - try this exploit."
"Wow - that's just what I was thinking of trying!"

It's also easy enough for nodes in the Matrix to operate on some basic rules and say - "hey, I have an army marching through here, lets alert someone or throttle this traffic". The Matrix was designed to be smart and to resist the sort of virus attacks that devastated the world. It's most emphatically not the Internet of today. And the RAW allow nodes this sort of intelligence so you can do it without problem.

I wrote a fluff piece on the Matrix a long time ago here which outlines some fluff justifications for this sort of resiliance towards the end.

If this poll were to be valid, it would need more options. As it is, it doesn't represent the range of viewpoints here on Dumpshock. We've already had quite a lot of these in the previous (long) thread on this. Polls aren't good for complex arguments like this, because people frequently vote and then read the following thread and can't change their mind. If you (Tarantula) are trying to sum up our previous debate on this and find out who "won" this poll wont really give you that.

EDIT: I null voted.

I go with the 1st one. There's nothing proving that it's right but contrary to what some people say, it can be explained and is consistent enough.

But that's an issue I'd like to see officially settled.

i think knasser covered most of the bases.

deploying a agent smith attack would be like a virtual 9/11. one would have the target, corp court grid overwatch division, and maybe some others to, descending on the event in short order.

its like massing the army at a border. its just not something you do without being noticed.

hell, isnt the one thing that keeps the corps in line the omega order? or basically a kind of corp MAD scenario? what would be the virtual equivalent?

yep, but then i think frank argues it as a pranksters weapon rather then a shadowrunners weapon. but then if matrix crimes where handled sharia style, i would think a lot of kids where a lot more careful...

I don't see anything "fundamentally flawed" here. Yeah, making DDoS attacks is possible. That's realistic. It's also a pretty unfeasible tactic in the context of most Shadowrun games, which is what the rules concern themselves with.

Your run-of-the-mill spam, botnets, and the like have been a problem for decades in 2070. Dealing with them in everyday life has been abstracted away by the rules. Using them as weapons in a hack probably has, too - with a little imagination, you could say that's part of what the "slow and sure" hacking rules are supposed to represent, for example. And if you counter that with "the rules shouldn't require you to have an imagination", go find another hobby .

What you're describing is a flood of thousands of agents, each running from its own dedicated commlink. A commlink capable of running the sort of agent a PC hacker is likely to employ will cost a couple thousand nuyen, at the very least. Want to send nine thousand "Agent Smiths"? Got eighteen million lying around?

And if the target in question has made itself impervious from external assault by the simple expedient of not being online, how do you carry 9000 commlinks into a place like that?

Alright, so using over nine thousand agents will get you too easily noticed and you have to deal with the whole MAD scenario. What if you don't use that many, and instead use half a dozen copy-pasted agents using teamwork. You can simply have a good commlink connected to whatever system you want protected, continually scanning the databases for any hackers, just buying hits and auto-detecting intruders in less than a combat turn. The same goes for any kind of nonopposed Matrix test, as you don't need to use a hundred agents when a dozen will make you unstoppable.

Sure, you'll be auto-detected, but the only real way to stop you once you're found is to shut down the system, and that alone will cause problems. And if MAD is initiated from a dozen agents, then the world is entirely too unstable.

At that scale, how can any corp function without having everything be an isolated island where data's transferred by optical disk ferried by a heavily guarded messenger goon?

That is the only objective of DoS.

that may be. but it sounded like it was a general problem of the matrix rules...

And what about the technique of getting a high-grade commlink loaded with half a dozen agents using Analyze on their designated system? It's a cheap purchase for any corp to have each agent to be different, and so teamwork is even more likely to work and thus you will auto-detect any and every hacker who works on his own; making the whole concept of covert hacking pointless. I don't see how this can be reconcilled without house rules, and thus my choice of simply hand-waving multiple agents into nonexistance for action resolution.

My house-rule reply to the "Agent Smith Problem:"

Agents are not treated as a simple program. In fact, they require special hardware that is included in the price of the purchase (usually a small box, no larger than a commlink). Corporations can buy much bigger boxes that they stick somewhere in their server farm (probably contains multiple high-quality agents).

The net result, of course, is that agents are treated as independent entities that don't have fluctuating stats, and if you want more than one, you just have to fork over the money. Corps got it, so they can afford to run dozens of agents (balanced by the fact that they have quite a few more nodes in their flock to tend) but players are usually strapped for cash.

The most agents I'd like to see (in terms of style and avoiding massive dice-induced headaches) in one hacker's hands would be 2-3. By their nature, agents require some serious processing power, and the idea of simply making copies ad infinitum until one side or another buckles seems to me to gut the more subtle aspects of the hacking rules.

Oh how I love that sweet, sweet zombie Jesus.

Yes but they are designed to act this way. You don't need an agent to connect to a webserver/ftp server/file server and so on. If you need to do that, a script is far enough.

Actually it just states that an agent can move from node to node and that the attributes of the agent will depend on the attribute of the node they're on.

But it doesn't state if the agent can act on another node than the node it is loaded on.
You may find it implies something... But I don't think rules should just imply on such important topics.

yes it dawned on me as i hit post that it can be interpreted both ways thanks to the use of "may". it could be that the writer had in his mind that different nodes would have different response rating, or that the agent didnt have to move.

something tells me there was a lot of text cut to fit the matrix rules into the book...

On a related note...

If programs are so difficult to copy, how is it that you can load a copy of a program you have onto an Agent and send him off to God-knows-where, yet still have access to the program yourself? And what's stopping you from performing this function with anyone else, effectively giving copies of the program away? I mean, if the Agent has everything it needs in the program loaded into itself to the point where it can leave your commlink completely, uhm, how's that any different from copying it?

The agent isn't copying the program, its loading it into memory and running it. Which takes a complex action. SR4, 227, "In order to activate a program, the user must first have the program available (either on his commlink or on one of his networked devices). The program must then be loaded into the persona’s memory and run, which takes a Complex Action. Running too many programs at once may affect Response (see p. 212)."

Once its loaded, you can move from node to node without issue. (since theres nothing that says you can't).

Which is my point.

What's the difference between copying a program and "loading" a copy of it into an agent or persona? It apparently has everything it needs from the program to go on independantly from its node of origin, so what's the difference between the two? And why can't you just copy the "loaded" copy to another agent or persona?

Fancy yet-unthought of copy protection? Good enough for me.

Except I imagine most nodes would allow guest accounts, for things like browsing a restaraunt menu, or looking at the store directory of a grocery store.

thats what arrows are for

Tarantula, is your initial assumption that Agents can only work differently then the Hacker Persona does in the first place?

Yes, they are not a hackers persona, and there is no text stating that they work the same.